HOD, Department of Computer Application & Assistant Professor - Department of Computer Engineering, B.H.Gardi College of Engineering & Technology, Gujarat, India
Authors - Md Mahmudul Hoque, Md Kawser Islam, Md. Mamunur Rahman Moon, Abdullah Rakib Akand, Md. Hadi Al-amin, H.M. Azrof Abstract - The automatic recognition of virus particles in transmission electron microscopy (TEM) images remains a demanding task, primarily owing to strong inter-class similarity, scale variability, and pronounced class imbalance. In this study, several convolutional neural networks and transformer-based architectures were comparatively evaluated for the classification of 22 virus categories using the TEM virus dataset. All models were trained under identical preprocessing and optimization conditions, and imbalance effects were mitigated through a weighted crossentropy formulation. Performance was quantified using overall accuracy together with macro-averaged precision, recall, and F1 score. Among standalone models, the Swin Transformer achieved the highest accuracy (0.8831) and macro-F1 score (0.8444), followed by DeiT (accuracy 0.8669). Convolutional architectures exhibited comparatively lower balanced performance, with ResNet50 demonstrating substantial degradation (accuracy 0.5887) under imbalanced conditions. To exploit complementary representational properties, decision-level hybrid strategies were implemented. The performance-weighted hybrid attained an accuracy of 0.8831 and the highest macro-F1 score (0.8528), slightly surpassing the equal-weight hybrid configuration. These observations indicate that architectural heterogeneity contributes to improved inter-class balance without sacrificing overall predictive accuracy. Future work may explore scale-aware representations, feature-level fusion mechanisms, and expanded TEM datasets to further enhance robustness and generalization in virus identification tasks.
Authors - SunilKumar Ketineni, Preethi Kandukuri, Hruthik Sreeramaneni, Vivek Bojjagani Abstract - Phishing continues to pose a serious threat to digital security by ex ploiting human vulnerabilities to steal confidential data through deceptive online interactions. Traditional detection methods often fall short in identifying advanced phishing strategies. This survey presents a comprehensive overview of phishing detection techniques, with a strong focus on modern, multi-layered machine learning and deep learning-based solutions. The proposed layered framework includes four key stages: data collection and preparation, model training, detection and prediction, and explainability. In the first layer, email, URL, and metadata are collected and preprocessed for feature extraction. The second layer involves model training using both machine learning classifiers such as Random Forest, SVM, Naïve Bayes, and KNN and deep learning archi tectures like CNN, RNN, and LSTM. These models feed into the third layer where phishing is detected and classified. Finally, the fourth layer integrates Explainable AI (XAI) methods like LIME, SHAP, and Anchors to enhance model transparency and interpretability. This survey evaluates the effectiveness and limitations of each layer and highlights the need for explainable, scalable, and adaptive phishing detection systems.
Authors - K.Poorani, K Karan, R Seenivasan, V Ramkumar Abstract - Older email detection technologies have struggled to accurately iden tify malicious emails in the face of the latest techniques attackers use to compro mise victims. While modern solutions perform well in detecting malicious emails, they are not completely foolproof. As a result, malicious emails can still reach a user’s mailbox, necessitating measures to reduce potential harm. This study suggests transforming the decision-making processes of recent algorithms into a white-box model, enabling transparency in decision-making through Ex plainable AI. This is achieved by having the proposed model compute confidence level scores for each email, which users can use to exercise caution if a malicious email slips into their inbox.
Authors - Nazura Javed, Rida Javed Kutty, Muralidhara B L Abstract - The increasing availability of online information has made it easier to access diverse sources, but it has also introduced challenges in verifying the reliability and consistency of content. Conflicting statements across different sources often contribute to misinformation and make it difficult to establish factual accuracy. This study focuses on the problem of cross-document contradiction and inconsistency detection as a step toward improving fact verification in textual data. A two-stage pipeline is proposed in which semantically related sentence pairs are first retrieved from documents discussing the same event and then analyzed using Natural Language Inference (NLI) techniques to determine whether they express contradictory information. In contrast to conventional sentence-level contradiction detection, the proposed approach emphasizes document-level comparison to identify inconsistencies across independent sources. Two pre-trained transformer models, DistilBERT (DistilBERT-base-uncased) and RoBERTa (RoBERTa-base), are used for contradiction classification. The approach is evaluated on the SNLI dataset and the PHEME Rumor Dataset, which are widely used benchmarks for NLI and misinformation research. Experimental results show accuracies of 94.50% (F1 score 94.50%) on SNLI and 92.39% (F1-score 92.31%) on PHEME, indicating that the proposed framework is effective in identifying contradictions and supporting cross-document fact validation.
Authors - B.Purnachandra Rao, Gaurang Jinka Abstract - Distributed systems rely on data replication across multiple nodes to ensure high availability, fault tolerance, and scalability. While replication improves system reliability, it also introduces temporary inconsistencies between primary and replica nodes during data propagation. This phenomenon, commonly referred to as consistency drift, occurs when distributed nodes maintain slightly different states before synchronization is completed. As distributed infrastructures grow in scale and complexity, consistency drift becomes increasingly significant due to network latency, workload variability, and communication overhead between nodes. Traditional synchronization mechanisms typically rely on static replication intervals or fixed update propagation strategies that do not adapt effectively to dynamic system conditions. Such approaches may allow drift to accumulate before synchronization occurs, resulting in delayed consistency and inefficient resource utilization. Managing consistency drift therefore becomes a critical challenge in distributed computing environments where maintaining accurate and synchronized data states is essential. This research addresses the problem of consistency drift in distributed systems by examining the factors that contribute to state divergence among nodes and exploring mechanisms for dynamic drift management. The proposed framework focuses on monitoring system behavior, including workload intensity, network latency, and node communication patterns, to regulate synchronization behavior more effectively. By enabling adaptive synchronization strategies that respond to real time system conditions, the framework aims to reduce drift accumulation and improve overall data consistency across distributed clusters. Effective management of consistency drift ultimately enhances system reliability, operational stability, and performance in modern distributed computing platforms operating under dynamic workloads.
Authors - Suganya Moorthy, Jayakumar Kaliappan Abstract - Internet of Things (IoT) networks have grown really fast, which has increased the attack surface of cyber attacks by a big mar gin. However, the severely limited computational resources, the hetero geneous architecture, and incomplete or decentralized communications make the IoT environments very susceptible to intrusion attacks, in cluding Distributed Denial of Service (DDoS), spoofing, botnets, and data exfiltration attacks. Older signature-based intrusion detection sys tem (IDS) is not effective in detecting zero-day and dynamic threats. The paper will present a new machine learning-based intrusion detection system, which was developed with IoT networks in mind. The design proposed combines the characteristics of feature search, feature detec tion, and group classification model in order to increase the accuracy of detection as well as reduce the number of computations. Benchmark IoT intrusion datasets that have undergone experimental evaluations prove to be more effective in detection accuracy, false positive rates and scaling than the traditional IDS frameworks. Practical constraints that include the computational overhead of resource-constrained IoT devices, imbal ance of the dataset, and interpretability of the model are addressed. The directions of future research are lightweight federated learning systems, explainable AI system incorporations, and real-time adaptive threat in telligence systems to build better resiliencies of IoT security.
Authors - Konstantina Karathanasopoulou, Ioannis Vondikakis, Dimitris Georgiadis, George Dimitrakopoulos Abstract - Digital signatures are fundamental public-key cryptographic primitives used for message authentication and integrity. A message’s recipient must be able to validate that it comes from the reported sender and hasn’t been altered by anybody else. Pairing-based cryptography provides elegant and efficient mechanisms for constructing compact dig ital signature schemes. Inspired by isogeny structures on elliptic curves, we present a pairing-based digital signature system in this study. Our construction targets classical security settings and is analyzed under standard computational hardness assumptions related to bilinear groups and isogeny-based mappings. We demonstrate that the proposed ap proach attains “existential unforgeability under adaptive chosen-message attacks (UF-CMA)” within the random oracle model and address the construction’s soundness and security. Moreover, the scheme offers com pact public key and signature sizes, making it suitable for lightweight cryptographic applications.
Authors - Nirmaladevi J, Kanishka R, Kirthiga B, Lathikasri T R, Ranjani Shree R S Abstract - The vast implementation of cloud computing has uplifted the modern IT practices by improving scalability, flexibility, and budget efficiency. In contrast, there has been an increase in energy consumption, which results in carbon emissions. This happens because of overusage, overconsumption, overprovisioning, unused capacity, and inefficient data center management. These days, data centers act as the sole contributor to global greenhouse gas (GHG) emissions; therefore, sustainable cloud operations are essential in addressing this challenge. GreenOps, or green operations, defines the cloud deployment and operational practices that take place but also considers the environmental impact; it depicts energy-efficient infrastructure design, optimized resource usage, virtualization, and the integration of renewable energy resources. This survey presents a summary of green cloud computing, including the current trends, challenges, energy-aware scheduling algorithms, and optimization techniques for obtaining energy-efficient cloud deployment.
Authors - Pranaav Contractor, Sanika Ajgaonkar, Nishanth Ravichandran, Satishkumar Chavan Abstract - This paper examines the interplay between demographic factors and a newly developed behav ioral construct—modern investment curiosity—and how these elements collectively shape finan cial behaviors among higher education faculty. Drawing from survey responses of 145 educators situated in Kollam District, Kerala, India, the study applies descriptive statistical techniques alongside chi-square tests to evaluate four research hypotheses. The data reveals a predominantly risk-averse financial posture among participants, with post-retirement security ranking as the foremost financial goal and bank deposits serving as the dominant investment channel. Statistical testing shows no meaningful relationships between saving patterns and either household size or disability status. A statistically significant positive association emerges between investment cu riosity and ownership of equity or mutual fund products (χ² = 8.40, p < 0.01). Additionally, mar ital status demonstrates a significant relationship with investment curiosity (χ² = 5.28, p < 0.05), where unmarried faculty report higher curiosity levels. These observations are consistent with established frameworks including the Life-Cycle Hypothesis and the Theory of Planned Behav ior, positioning investment curiosity as a relevant psychological factor in financial decision-mak ing. The paper offers practical suggestions for institutional programming and identifies avenues for subsequent scholarly inquiry.
Authors - B.Purnachandra Rao, Gaurang Jinka Abstract - Distributed systems rely on data replication to ensure availability, fault tolerance, and scalability across multiple nodes in modern cloud environments. Replication enables systems to maintain continuity even when individual nodes fail or experience network disruptions. However, replication often introduces synchronization delays between primary and replica nodes, known as replication delay. These delays can cause temporary data inconsistency, stale reads, and increased response latency, degrading application performance and user experience. As infrastructures scale to larger clusters, communication overhead, network latency, and workload variability further amplify replication delays, making efficient synchronization increasingly challenging. Traditional replication mechanisms typically rely on static synchronization intervals or sequential update propagation strategies. These approaches fail to adapt to dynamic network conditions and fluctuating workloads, resulting in inefficient data propagation and delayed consistency across nodes. In large scale systems, such limitations may cause bottlenecks, reduced reliability, and inconsistent states during high workload periods or network congestion. Addressing replication delay is critical for maintaining reliability and consistency in distributed environments. Recent research emphasizes intelligent synchronization mechanisms capable of adapting to changing conditions. Adaptive synchronization strategies that monitor network latency, workload intensity, and node communication patterns offer improvements in replication efficiency. By enabling replication decisions that respond dynamically to system behavior, such approaches reduce synchronization delays and improve data consistency across clusters. Enhanced replication efficiency ultimately strengthens reliability, scalability, and operational performance in modern distributed computing platforms operating under variable workload conditions.
HOD, Department of Computer Application & Assistant Professor - Department of Computer Engineering, B.H.Gardi College of Engineering & Technology, Gujarat, India
Authors - Konstantina Rigou, George Dimitrakopoulos Abstract - The rapid adoption of Artificial Intelligence (AI) in high-impact domains (healthcare, finance, justice) creates an urgent need for sys tems that are legally compliant, explainable, ethical and transparent. Decision Support Systems (DSS) aim to assist managerial and professional decision-making, yet few works translate legal and ethical principles into concrete technical design constraints for explainable AI (XAI). This paper proposes a Legal Explainability Framework (LEF) that maps legal obligations (General Data Protection Regulation, European Union Artificial Intelligence Act) and ethical principles to measurable XAI requirements and implementation steps, and demonstrates the approach with a prototype using an open legal dataset derived from judgments of the European Court of Human Rights (ECtHR). The results show that legally compliant XAI is not merely a normative aspiration, but a technically feasible and practically implementable design paradigm.
Authors - P.Pandiaraja, N.Shiva Kumar, B.Vishnu Vardhan, C.Sevarathi, Charles Prabu V, S.Jagan Abstract - Retrieval-Augmented Generation (RAG) chatbots represent a significant advancement in intelligent conversational systems, grounded in the prin-ciples of natural communication, accuracy, and reliability. Traditional chatbots are constrained by pre-trained knowledge or rule-based responses, limiting their effectiveness in dynamic and complex real-world scenarios. RAG-based systems integrate information retrieval mechanisms with sophisticated language generation models to identify relevant knowledge in real time and produce contextually appropriate responses. The proposed system employs sentence-transformers (all-MiniLM-L6-v2) for dense vector embeddings and FAISS as the vector data-base backend, enabling fast and semantically accurate document retrieval. Ex-perimen- tal results demonstrate a mean retrieval accuracy of 87.4%, an average response latency of 1.3 s, and a user satisfaction score of 4.2 out of 5, confirm-ing the system’s readiness for real-world deployment.
Authors - Manjula K, Vijayarekha K, Venkatraman B Abstract - The fabrication of components across various industries is accom plished through welding. Although welding has been practiced for more than a hundred years, defects may still occur during the welding process. Thus, indus trial standards require welded joints to be inspected and evaluated to ensure their quality and reliability. Conventional ultrasonic testing (UT) has long been widely used in industry for detecting and evaluating defects in weld specimens. Over the last few decades, advances in sensor technology and signal analysis techniques have significantly advanced ultrasonic testing methods. Advanced methods, such as Time Of Flight Diffraction (TOFD), are more likely to detect linear defects. However, one of the major challenges in applying TOFD to the inspection of austenitic stainless steel (ASS) weldments is noise in the signals. Various signal processing approaches have been developed to suppress such noise, each with its own advantages and limitations. In this work, the focus is placed on the applica tion of multi-level discrete wavelet transform (DWT) decompositions with ‘n’- order wavelet filters for de-noising ultrasonic TOFD A-scan signals. The results show that this approach achieves greater improvement in signal-to-noise ratio (SNR) while requiring less computational time.
Authors - Likhitha Ragha Ramya Nakka, Anuradha Andra, Appalaswami Ravada, Vinay Kumar Pamula Abstract - This study uses Roland Barthes' semiotic approach to analyze how meaning is represented in HMNS' Untitled Humans ad on Instagram Reels. Understanding how storytelling campaigns create and communicate meaning has become crucial for successful digital marketing as social media plays a big-ger role in brand communication strategies. This study examines a selection of Instagram Reels content from the official Instagram @hmns account using a qualitative-descriptive methodology, emphasizing how text, sound, and visual components interact to provide multiple interpretations. The study methodically sign how everyday occurrences, human relationships, and nature scenery are turned into symbolic representations of authenticity, freedom, and personal identity using Roland Barthes' three-level semiotic framework: denotation, connotation, and myth. Direct observation and content documentation of Reels recordings are used for data gathering, and triangulation is used for analysis to guarantee validity and thoroughness. Results show that by creating an existential story that prioritizes closeness, introspection, and human connection, the campaign goes beyond traditional product advertising. Authentic, unconstructed life imagery is presented at the denotative level, visual and musical elements evoke emotion and personal memory at the connotative level, and perfume, rather than being a commercial product, becomes a symbol of emotional intimacy and identity exploration at the mythic level.
Authors - Deepak Mane, Siddhi Dhamal, Shivam Devkar, Divit Maheshwari, Riddhi Kaulage, Diya Nair, Deepak R. More Abstract - The evaluation of handwritten answers sheet has so many challenges since from many years due to variability in handwriting, linguistic barrier and personal bias. This is very time-consuming method and inconsistent method which highlights the need for automated subjective answers evaluation. Here, proposed automated handwritten answers evaluation system uses TrOCR based handwritten answer detection, NLTK tokenization, WordNet lemmatization and semantic similarity check between teacher’s and student’s answer based on meaning. This advanced multi-model system overcomes traditional keyword matching technique and improves contextual accuracy. This system also overcomes traditional manual checking and results in fast evaluation. The system promotes the fairness, fast and accurate processing. Moreover, the suggested framework removes human fatigue, encourages fair grading, and offers a solution that can be used for large-scale academic tests. The results show that this automated method not only works like a human brain but also makes the evaluation process more fair and open.0
Authors - Deepak Mane, Deepak R. More, Arya Kale, Ravina Jagtap , Soumya Dubewar , Diya Nair Abstract - Timely detection of crop diseases is essential to ensuring high agricultural produc- tivity; thus, early and accurate detection has always been a priority for the farmers. So we pro- posed a deep learning based framework that classifies the condition of basil leaves in three cat- egories - wilting, infection by mildew and healthy - through an EfficientNet-B0 convolutional neural network fine-tuned using transfer learning. We leverage a curated dataset of 1,442 plant images available at the Roboflow platform, splitting the dataset into 70% training, 20% valida- tion and 10% testing. Transfer learning was used where we started EfficientNet-B0 with weights learned on large scale ImageNet pretraining. Training was done in two stages: first the whole model was trained with the backbone frozen and only the newly added classification head being trained, followed by unfreeze the last 100 layers and perform fine-tuning to the domain. Leaf orientation and illumination variability were treated by a group of data augmentation methods including random horizontal flipping, rotational transforms, zoom perturbations, and contrast adjustments. The proposed system achieved a remarkable result with high generalization of 96.6% training accuracy and 97.8% test accuracy. The detailed analysis of the confusion matrix and the ROC-AUC curves corroborate faithful multi-class discrimination. A Streamlit-based web interface was also developed to facilitate live inference, farmers and agronomists are now able to make immediate predictions of the disease with confidence estimates. The results showed that the well optimized EfficientNet-B0 model can be a feasible and scalable solution for automated monitoring of crop diseases in the context of smart agriculture.0
Authors - Vinodkumar Bhutnal, Prajwal Vijay Sonawane, Om Vinod Chaudhari, Avinash Golande, Mohit Ashok Tajane, Sujal Kishor Papdeja Abstract - There is no more pressing issue in modern cities, industries, and public venues than nighttime security, as the conventional approach of patrolling in-person only works well until fatigue and coverage become challenges, when humanity and human error become a finite issue that requires short delay interruptions. Urbanization, increased crime rates, and the inadequacy of current traditional patrolling to provide a sufficient security posture have led to the proposal of an Intelligent Night Patrolling System that uses edge-cloud frameworks, IoT-enabled CCTV camera technology, and artificial intelligence video analytics to significantly reduce the presence gap. This system will provide continuous, real-time proactive surveillance of locations and even be equipped with advanced deep learning models like Cummings Neural Networks (CNNs) and Long Short term Memory (LSTM) to detect suspicious activity, anomalies, intrusions, and violent types of activities. This research introduces the concept of Night Patrolling System designed to assist security personnel during night surveillance.The proposed system achieves an estimated accuraxy of over 90% with a reduced latency , demonstarting it’s effectiveness for a real time survillence applications.
Authors - Deepak T. Mane, Deepak R. More, Gopal D. Upadhye, Rucha C. Samant, Hemlata U. Karne, Suraksha Suryawanshi, Prem Borse Abstract - Efficient vehicle type classification is vital for intelligent transportation systems, traffic monitoring, and urban mobility planning. This paper presents a Real-time Multimodal Vehicle Type Classification System that leverages both visual and acoustic data to identify and categorize vehicles such as cars, buses, trucks, and motorcycles from live video streams. The proposed system integrates CNN-based and Transformer- based models for feature extraction across modalities, enhancing detection robustness under diverse lighting, weather, and traffic conditions. A lightweight preprocessing pipeline performs synchronized frame extraction, audio segmentation, and feature fusion while ensuring minimal latency in real-time environments. The proposed multimodal architecture combines late fusion of visual and audio features to enhance the reliability of classification when either modality is suffering from low visibility or occlusion. Experimental evaluations demonstrate that the proposed framework achieves a classification accuracy of 96.2% at 28 fps, outperforming unimodal baselines with real-time efficiency. This system is deployable for intelligent traffic surveillance, automated tolling, and urban safety analytics.
Authors - Shwetha Ramadas, Krutthika Hirebasur Krishnappa, Sudhir Trivedi Abstract - Methane (CH4) emission from rice paddies is a significant source of greenhouse gas emissions from agriculture. Currently, most models for methane prediction from rice paddies depend on collecting field data and sending it to a server. In this new paradigm, several privacy concerns arise, model scalability is restricted, and a large number of data points are exposed to the attacker. This paper addresses all privacy con cerns by providing an edge-based solution for modeling methane emis sions from rice paddies that leverages data from edge sensors at respec tive locations, while keeping individual sensor data private. The method employs different machine learning (ML) algorithms, including Linear Regression, Random Forest, XGBoost, and a Feedforward Neural Net work (FNN), implemented using TensorFlow Federated (TFF) in both centralized and federated learning (FL) frameworks. The FL-based FNN achieved an R2 score of 0.91, which was superior to both centralized classical and centralized FL models, especially for highly non-IID client side data distributions in sensor datasets. In summary, this paper extends the current literature on modeling methane emissions from rice paddies and provides a comprehensive evaluation of our proposed FL system ar chitecture, an in-depth discussion of the communication resources re quired for FL implementation, and an examination of the effects of abla tion studies on clients’ data heterogeneity. Therefore, the proposed FL approach is efficient and scalable, enabling safe, privacy-preserving modeling of methane emissions from rice paddies to effectively imple ment Climate Smart Agriculture (CSA) and mitigate global warming while supporting sustainable rice cultivation.
Authors - P. Pandiaraja, P.Krishna Kishore, E. Ganesh, C. Selvarathi, Charles Prabu V, S. Jagan Abstract - Large Language Models have facilitated the development of sophist i-cated smart platforms that are actively leveraged in the provision of financialservices to various classes of customers. This advancement has enabled peopleto obtain individual financial advice. This paper presents a framework for buil d-ing a financial chatbot that incorporates Retrieval Augmented Generation(RAG) technology and several SQL agents to improve reliability. The proposedapproach addresses five fundamental challenges in financial artificial inte ll igence: eradicating hallucinations, obtaining up to date information, utilising u s-er facts to tailor individual suggestions, safeguarding user privacy, and provi d-ing clear explanations. RAG is used to retrieve verified financial knowledge,while SQL agen ts query databases to produce accurate outputs. The solutionprovides advisory responses that are relevant to users and protect sensitive i n-formation through a zero trust security architecture. The system architecture i n-corporates multiple validation check points and is dynamically configured tomeet individual user requirements. Experimental results demonstrate a 96.2%accuracy rate in handling financial queries with a 3.8% error rate and a mean r e-sponse time of 1.5 seconds, outperforming comparable solutio ns. The proposedarchitecture establishes a reliable baseline for financial professionals seekingdependable advisory services.
Authors - Mohd. Zuhaib Ahmed, Akash Priya, Deepti Chopra, Pankaj Kumar Abstract - Effective landing and take-off (LTO) decision-making in mil itary aviation is critically dependent on airfield serviceability and pre vailing weather conditions. A fundamental challenge is the absence of structured expert pilot decision logs, as such data are operationally sen sitive and access-restricted. This work presents a replicable methodolog ical framework for developing machine learning-based decision support systems in domains where operational data are scarce or classified. The pipeline encompasses synthetic data forged using correlated Monte Carlo sampling, constrained by location-specific geographic, seasonal, and ter rain parameters across ten Indian Air Force (IAF) stations, yielding ap proximately 60,000 simulated operational scenarios. The dataset is gen erated within domain-constrained operational bounds to ensure physi cal plausibility. A rule-based expert classification system assigns opera tional status as Green (Safe), Orange (Caution), or Red (Unsafe); four ML algorithms are subsequently evaluated: Logistic Regression, Naïve Bayes, Support Vector Machines, and Decision Trees. The Decision Tree achieves the highest performance, with an accuracy of 0.983, an F1 score of 0.983, and a ROC-AUC of 0.984. The proposed framework supports two deployment pathways: the rule engine as a deterministic automa tion tool for standard clearances, and the ML model as the inference core of a real-time Human-in-Loop (HIL) expert system requiring opera tor authorisation at every decision. As expert pilot decision logs become available, the system may be progressively elevated to a fully adaptive expert system.
Authors - Ritesh Kumar Verma, Preethiya T Abstract - Contemporary customer support systems require processing a massive number of user queries with low latency and high semantic relevance. Rule-based systems fail to capture context, while fully LLM-based systems are computation ally expensive and suffer from high latency. This paper introduces an adaptive AI-assisted customer support automation system using an optimized Retrieval Augmented Generation (RAG) model. The proposed system combines Azure OpenAI embeddings, FAISS-based vector search, selective Cross-Encoder re ranking, and a Learning-to-Rank (LambdaMART) model for adaptive score fu sion. Unlike vanilla RAG models, the proposed system adaptively re-ranks only the top-k retrieved candidates, trading off ranking precision and latency. Experi ments were carried out on a 1,30,000-sample e-commerce customer support da taset with query-response pairs annotated with intent labels. Compared to rule based retrieval, embedding+FAISS, and vanilla RAG models, the proposed hybrid system showed improved top-1 retrieval precision with a concurrent reduc tion in end-to-end latency from 0.414s to 0.365s (≈11.8% relative improvement). The LambdaMART model adaptively learned weights from FAISS and Cross Encoder scores, improving ranking robustness and eliminating misranked top re sponses. The system was implemented on Azure Machine Learning with a cloud scale pipeline and interactive Streamlit web interface, showcasing the cost-effec tive inference capabilities of the proposed system via selective re-ranking.
Authors - Abdelrahman El Antably, Ali Hamdi, Ammar Mohamed Abstract - Large Language Models (LLMs) frequently generate plausi ble but incorrect information, known as hallucinations. Detecting these errors at a fine-grained level is crucial, especially for morphologically rich languages like Arabic with limited resources. We introduce BAL ANCE:Bi-perspective Analysis for LLM Accuracy via coNsensus ChEck ing, a novel dual-judge framework for token-level hallucination detection in Arabic LLM outputs. Our six-module pipeline features context filtra tion, argument decomposition, and distinct strict and lenient LLM-based judges. A consensus coordinator then synthesizes their verdicts, and a span annotator precisely localizes errors. Evaluated on the Arabic sub set of the SemEval-2025 MuSHROOM benchmark, BALANCE achieved an Intersection over Union (IoU) score of 72.87%. This significantly outperforms the task’s winning system by approximately 8.76% rela tive improvement and consistently surpasses zero-shot baselines across various LLMs by up to 39.80 percentage points.
Authors - Duy Pham, Tung-Duong Le-Duc, Anh-Tai Pham-Nguyen, Trung Nguyen Mai, Long Nguyen, Dien Dinh Abstract - Multimodal knowledge graphs improve structured knowledge representation and tasks such as cross-graph entity alignment. However, most benchmarks focus on resource-rich languages and assume dense relational structures and balanced attributes. Low-resource languages like Vietnamese pose additional challenges, including structural sparsity, attribute asymmetry, and modality noise. To address this gap, we in troduce DBWiki-VN15K, the first large-scale Vietnamese multimodal knowledge graph dataset for entity alignment. Built from Wikidata and DBpedia, it contains 15,000 aligned entity pairs with relational triples, lo calized numerical attributes, and visual modalities. The dataset provides both word-segmented and unsegmented text to support different linguis tic processing approaches. Experiments with state-of-the-art multimodal entity alignment models reveal that structure-guided multimodal fusion and dynamic modality weighting are more robust to sparse and noisy features. Additionally, unsegmented subword tokenization better han dles cross-graph translation inconsistencies than strict Vietnamese word segmentation. DBWiki-VN15K offers a realistic benchmark for studying multilingual and multimodal knowledge fusion. Our dataset is available at: https://github.com/Tim50c/DBWiki-VN15K.
Authors - Ritesh Jawarkar, Reena Satpute, Sudhir Agarmore Abstract - Because sleep problems can influence the health of a person and his/her quality of life, such diagnosis and treatment relies on specific classification. Even though single deep learning and machine learning models have shown their potential, they are limited by overfitting and bias in the model. In order to solve these issues, the current research proposes the expansion of the ensemble learning-based sleep disorder classification through the inclusion of machine learning model predictions. A voting classifier enhances the optimization base classifier outputs in terms of robustness and classification accuracy. According to Sleep Health and Lifestyle Dataset, the ensemble method has 97.3 percent accuracy with individual models. The interface is designed as a Flask-based web interface that allows user authentication to increase user interaction and usage of the system on a real-time basis. Suggested extension ensures the reliable, accurate and easy-to-use automated sleep problem diagnosis.
Authors - Aman Kumar, Mary Subaja christo Abstract - Content Delivery Networks (CDNs) play an essential role in enhancing the content delivery speed by caching frequently requested data in edge servers distributed across geographical regions. Traditional CDNs utilize rule-based policy and machine learning approaches for optimizing the cache. Machine learning is performed centrally, and the cache optimization is performed using the traffic logs collected by the central server. Although the use of central learning approaches is beneficial, it poses certain limitations, including data privacy and high communication cost. The central learning approach aggregates raw data, which poses data privacy issues. This paper proposes an architecture for secure federated learning, which is utilized for cache hit prediction in CDNs. The proposed architecture is evaluated using a synthetic dataset containing 1,30,548 records, and the features include temporal and network features. The proposed architecture is compared with the traditional central learning approach, and the results reveal that the secure federated learning model achieves an accuracy of 70.15%, which is comparable to the central learning approach. The proposed architecture is found to reduce data privacy exposure by 30%.
Authors - Bambang Marsudi Salim, Hudan Studiawan, Baskoro Adi Pratomo Abstract - Digital forensic investigations face a growing threat from sophisticated log tampering, in which adversaries delete or modify computer event logs to conceal evidence of criminal activity. This paper presents an empirical comparison of A Search and Iterative Deepening A* (IDA*) for reconstructing falsified computer event logs, extending the previous bipartite graph framework. Three log artefacts were constructed from the public forensic timeline dataset: an original computer log, a trusted ISP log, and a deliberately falsified log containing 15 strategically deleted events. To address timestamp heterogeneity arising from different system and ISP browser log parsers, a window-based matching strategy is introduced. Experiments conducted across maximal consecutive event sequences (MCES) demonstrate that IDA* consistently explores fewer nodes than A*. Anomaly detection identified 60.7% of browser events as uncorroborated by ISP records, achieving 60.0% recall on the 15 deliberately deleted events.
Authors - Akshay Ladha, Supraja P Abstract - Twitter social media platforms have become the primary means of communication for customer support, requiring rapid, accurate, and scalable response solutions. Conventional customer support mechanisms are primarily manual and inefficient in handling large volumes of real-time interactions. This paper presents an AI-Assisted Twitter Support System that combines deep learning with distributed streaming engines to automate real-time customer interactions. The system design utilizes Apache Kafka for tweet streaming, Apache Spark Streaming for distributed processing, and Long Short-Term Memory (LSTM) networks for sentiment analysis and multi-class complaint classification. A confidence-aware decision-making module is used to ensure that automated responses are produced only when the prediction confidence level exceeds certain thresholds, thus avoiding potential miscommunications. The system was trained and tested on the Kaggle Airline Sentiment dataset (1,46,400 tweets) with three sentiment classes and eight complaint categories. The sentiment analysis model attained an accuracy of 85.2% (F1-score of 0.846), and the complaint classification model attained an accuracy of 80.5% (F1-score of 0.792). The complete pipeline maintained an average latency of 2.9 seconds with a maximum processing rate of 2500 tweets per minute.
Authors - Pravitha N R, Sreelakshmi S R, Valsalachandran K, Savithri S Abstract - The rapid expansion of digital services has significantly increased the collection and processing of personal data through online platforms such as e-commerce systems, social media applications, and digital payment services. To regulate the use of personal information, governments worldwide have introduced data protection regulations such as the General Data Protection Regulation (GDPR), the Digital Personal Data Protection Act (DPDPA), and the California Consumer Privacy Act (CCPA). Organizations publish privacy policies to inform users about their data practices; however, these policies are often lengthy, complex, and difficult for users to understand. Consequently, users frequently accept privacy policies without fully reviewing how their personal data is collected, processed, and shared. Recent research has explored automated approaches for privacy policy analysis using artificial intelligence techniques, including machine learning, natural language processing, and large language models. Retrieval-Augmented Generation (RAG) has further enhanced compliance evaluation by linking policy statements with relevant regulatory clauses. Despite these advancements, challenges remain, such as the lack of standardised datasets, limited explainability of AI decisions, dependence on prompt design, and insufficient validation with regulatory experts. This paper discusses future research directions in AI-driven privacy policy compliance analysis and highlights emerging opportunities for improving regulatory compliance assessment, user privacy protection, and transparent privacy governance in digital ecosystems.
Authors - Ayushi Raj, Malathy C Abstract - The rapid growth of sensitive data requires backup systems that are both storage-efficient and risk-aware. Traditional backup approaches rely on static policies that ignore temporal changes, data sensitivity, and redundancy, leading to inefficient storage use and higher risk exposure. This work proposes a risk-adaptive backup optimization framework integrating temporal modelling, sensitivity-aware deduplication, and online learning. The system reconstructs data evolution using intrinsic timestamps and quantifies data criticality through continuous sensitivity scoring. A unified risk model combines sensitivity, change intensity, and exposure over time to determine backup urgency. An online rein forcement learning agent dynamically optimizes backup decisions based on evolving data patterns. The framework applies secure, sensitivity-based dedupli cation to reduce redundancy while preserving privacy. Operating in a read-only, metadata-driven manner, it ensures compliance with strict data governance re quirements. By decoupling decision logic from storage, the system supports hy brid cloud environments. Experimental results show reduced storage costs and controlled risk, demonstrating its effectiveness for scalable, intelligent data pro tection.
Authors - Nimisha K, Sridharan G, Kathiresh kumar K, Lohit S, Shyam Ganesh K Abstract - The rapid growth of sensitive data requires backup systems that are both storage-efficient and risk-aware. Traditional backup approaches rely on static policies that ignore temporal changes, data sensitivity, and redundancy, leading to inefficient storage use and higher risk exposure. This work proposes a risk-adaptive backup optimization framework integrating temporal modelling, sensitivity-aware deduplication, and online learning. The system reconstructs data evolution using intrinsic timestamps and quantifies data criticality through continuous sensitivity scoring. A unified risk model combines sensitivity, change intensity, and exposure over time to determine backup urgency. An online rein forcement learning agent dynamically optimizes backup decisions based on evolving data patterns. The framework applies secure, sensitivity-based dedupli cation to reduce redundancy while preserving privacy. Operating in a read-only, metadata-driven manner, it ensures compliance with strict data governance re quirements. By decoupling decision logic from storage, the system supports hy brid cloud environments. Experimental results show reduced storage costs and controlled risk, demonstrating its effectiveness for scalable, intelligent data pro tection.
Authors - Anirudh P, Nimisha K, Princy P Abstract - As technology advances, circuit complexity increases, integrated cir cuits become more prone to defects during manufacturing and operation. Conse quently, in order to ensure reliable operation, effective testing and stability eval uation of memory cells are essential. Static random-access memory plays a major role in modern digital systems due to its high-speed data access and efficient per formance. However, its reliable functioning is strongly influenced by device level parameters and supply voltage variations. In critical applications, even single fault occurrence may pose serious reliability issues, highlighting the need for ef ficient test methods. Extensive research has been carried out to investigate the static noise margin of SRAM cells. However, the influence of multiple defects has received relatively limited attention in existing literature. This study empha sizes the analysis of multiple defects because their occurrence becomes more fre quent in nano-meter technology regimes. Moreover, these defects can cause sig nificant fault behavior, potentially reducing the stability and reliability of SRAM cells. Multiple defects (Df3-Df3c) and (Df4-Df4c) are selected for analysis as they produce strong fault effects as they occur in the power supply and ground paths of the SRAM cell, which are critical for proper circuit operation. Any dis turbance along these conduction paths alters the effective operating voltage of the cross-coupled inverters and consequently affect the drive capability of the associated transistors. Moreover, the behavior of these defects is examined under various temperature conditions, supply voltages, and process corners in order to assess their overall effect on SRAM cell stability.
Authors - Sunil Jagannath Panchal, Gajanan Madhavrao Malwatkar Abstract - This research deals with the persistent challenges of document man agement in higher education institutions which focuses on the development of a digital support tool for Mariano Marcos State University (MMSU). Traditional paper-based systems and fragmented repositories often result in inefficiencies, duplication of work, and risks of data loss. The project adopted the Agile Devel opment methodology with emphasis on flexibility, collaboration, and iterative improvement. The d-T.R.A.I.L. system was built using JavaScript, PHP Laravel, HighCharts, and MySQL, integrating features such as tagging, repository man agement, granular access control, and collaborative modules like Teams. These functionalities were designed to streamline document organization, retrieval, and secure sharing across diverse academic and administrative units of the Univer sity. A User Acceptance Test (UAT) was conducted involving 70 participants from different MMSU offices that utilizes a Likert scale to measure satisfaction. Re sults yielded an overall mean score of 4.36 which was interpreted as Very Satis factory. High ratings were recorded for productivity, user-friendliness, and doc ument organization, while scalability received the lowest score which indicates an area for future enhancement of the system. The findings demonstrate that the system effectively improves workflow efficiency, accessibility, and accountabil ity, while aligning with national digital transformation policies.
Authors - Hileni Ihambo, Fungai Bhunu Shava, Gabriel Tuhafeni Nhinda Abstract - Fine-tuning large language models remains costly, and Parameter- Efficient Fine-Tuning (PEFT) techniques have emerged to make this process feasible on limited hardware. Among them, IA3 stands out for its extreme simplicity—it tunes only element-wise scaling vectors—but this design restricts the model to re-weighting features it already knows; it cannot form new ones. In this paper, we present SAMA (Spectral- Aware Minimal Adaptation), an extension of IA3 that introduces a single rank-1 update whose direction is derived from the pre-trained weights through Singular Value Decomposition. Each adapted layer gains only 4d extra parameters (3,072 for d=768), which is roughly one quarter of what LoRA requires at rank 8. We benchmark SAMA against five baselines—LoRA, DoRA, AdaLoRA, QLoRA, and IA3—across BERT, GPT-2, and Flan-T5 on twelve diverse NLP tasks under a low-resource constraint of 1,000 training samples per task. On the decoder-only GPT- 2, SAMA lowers perplexity by 26–34% relative to IA3 on both WikiText- 2 and Penn Treebank. On BERT’s RTE task, SAMA reaches 53.7% accuracy, surpassing IA3 (47.2%) and LoRA (52.6%) despite using 63% fewer trainable parameters than LoRA. We investigate this architecture dependence in detail and distil practical guidelines to help practitioners choose the right PEFT method for their setting.
Authors - M SANTHIYA, V KALAICHELVI Abstract - The wide use of machine learning in the field of medical imaging has caused concern with regard to patient information security, especially when mod els are being trained over multiple health care systems in a distributed manner. Centralized learning requires transferring raw patient data to a central server where there is an extreme risk of data breach and unauthorized access to patients' personal information. Violations of health care regulations (HIPAA and GDPR) can occur in a centralized system because of the transfer of patients' data. Feder ated Learning (FL) addresses these issues by allowing collaborative model de velopment on individual client devices. Therefore, the sensitive patient data will remain at its source institution. This paper provides a thorough comparative study of centralized learning and federated learning methods for detecting pneumonia utilizing chest X-rays from the publicly available Kaggle Chest X-Ray Pneumo nia dataset. Three architecture types (Support Vector Machine (SVM), Convolu tional Neural Network (CNN) and Long Short-Term Memory (LSTM)) were tested in both centralized and federated environments utilizing the FedAvg ag gregation method. Only the model weights were shared between the clients and the central server; therefore, patient data was maintained private through the en tire model training process. Experimental results demonstrated that federated learning produced superior performance than centralized learning for all three architectures (81.1%, 84.6%, and 82.7% for SVM, CNN and LSTM respec tively). The performance metrics for centralized learning were 76.6%, 76.3%, and 81.6%. This superior performance of FL is attributed to the inherent regular ization effect of local class-balancing within the federated clients that reduces the inherent class imbalance in the dataset. Overall, our research demonstrates that FL is not only a viable privacy-preserving solution to centralized training but offers improved generalization in the medical imaging domain with imbalanced classes and is a suitable solution for application in distributed health care envi ronments.
Authors - Vishruth B. Gowda, Sowmya T, Shreyas K, Megha J, Shreenidhi B S, Pranav Srinivas Abstract - Public administrations generate extensive administrative data through routine governance processes yet it is weakly based on verifiable evidence. This paper introduces a human-centric policy intelligence system based on execution-level administrative data for provision of accountable and evidence-based policy-making. The framework brings together governance-conscious data ingestion, cryptographic hash-based verification including permissioned blockchain systems to control the integrity of data, cross-domain data harmonisation to overcome administrative silos, and explainable machine learning models to create interpretable supporting insights. The framework is specifically meant as a human-in-the-loop system, maximizing policy foresight, administrative discretion, and accountability to the law. The validation with actual Mahatma Gandhi National Rural Employment Guarantee Act administrative data of the year 2022–2023 proves that the framework can be used to stress the implementation issues and regional inequalities without computerising policy-related decisions. The suggested solution is lightweight, scaled down to fit in the existing open-sector digital infrastructure.
Authors - Aprna Tripathi, Akhilesh Kumar Sharma, Avisikta Pal, Srikanth Prabhu, Ramakrishna Mundugar, Reet Ginotra Abstract - This paper presents a novel approach to identifying translation errors in Thai-English machine translation through the comparative analysis of multiple automatic evaluation metrics. Using a rank deviation methodology, we evaluate 350 Thai-English translations produced by seven contemporary systems provid ing online translations — including dedicated Machine Translation systems and large language models — across nine automatic evaluation metrics. By ranking translations within each metric and comparing individual metric rankings against the mean average rank, we identify translations that receive solitary punishment from a single metric, isolating these as candidates for manual error analysis. Our results demonstrate that individual metrics exhibit distinct sensitivity to specific error types, and that surface-level metrics retain diagnostic value along side advanced neural metrics. Neural metrics effectively identify meaning and adequacy errors, but surface-level metrics uniquely identify morphological vari ation, word order errors, preposition choice, and number formatting issues that neural metrics fail to penalize. The diversity of metric sensitivity is therefore an asset rather than an inconvenience, enabling a more complete characterization of translation error than any single metric can provide. This research supports the development of high-quality training data for MT fine-tuning by identifying the specific error types that individual metrics can and cannot detect and provides a repeatable diagnostic methodology applicable to other language pairs.
Authors - Bobby A. Eclarin, Mark Justine S. Cudapas Abstract - This research deals with the persistent challenges of document man agement in higher education institutions which focuses on the development of a digital support tool for Mariano Marcos State University (MMSU). Traditional paper-based systems and fragmented repositories often result in inefficiencies, duplication of work, and risks of data loss. The project adopted the Agile Devel opment methodology with emphasis on flexibility, collaboration, and iterative improvement. The d-T.R.A.I.L. system was built using JavaScript, PHP Laravel, HighCharts, and MySQL, integrating features such as tagging, repository man agement, granular access control, and collaborative modules like Teams. These functionalities were designed to streamline document organization, retrieval, and secure sharing across diverse academic and administrative units of the Univer sity. A User Acceptance Test (UAT) was conducted involving 70 participants from different MMSU offices that utilizes a Likert scale to measure satisfaction. Re sults yielded an overall mean score of 4.36 which was interpreted as Very Satis factory. High ratings were recorded for productivity, user-friendliness, and doc ument organization, while scalability received the lowest score which indicates an area for future enhancement of the system. The findings demonstrate that the system effectively improves workflow efficiency, accessibility, and accountabil ity, while aligning with national digital transformation policies.
Authors - Gauthaman S P, Paneer Thanu Swaroop C, Bagavathi Sivakumar P, Anantha Narayanan V Abstract - Psoriasis is a long-term inflammatory skin disease commonly identi fied by red plaques, scaling, and abnormal thickening of the epidermis. Reliable evaluation of disease severity is important for determining appropriate treatment options and for tracking patient response to therapy. In clinical practice, severity is often assessed using the Psoriasis Area and Severity Index (PASI). Although widely adopted, this method largely depends on visual examination and clinician judgment, which may lead to inconsistencies and observer-dependent variations. Recent developments in artificial intelligence and non-invasive dermatological imaging technologies provide opportunities for more objective and automated assessment of skin disorders. In this study, a novel framework is proposed for psoriasis severity evaluation that integrates skin biomechanical characteristics with deep hybrid learning mod els. Biomechanical attributes of the skin, including elasticity, stiffness, and vis coelastic behavior, are obtained through non-invasive measurement techniques and combined with visual information derived from dermatological images. The proposed system employs a hybrid deep learning architecture that incorporates convolutional neural networks (CNN) for image feature extraction along with machine learning classifiers for severity prediction. By jointly analyzing biome chanical and visual features, the framework aims to enhance the precision, con sistency, and reproducibility of psoriasis severity assessment. Experimental anal ysis indicates that the inclusion of biomechanical biomarkers alongside deep learning significantly improves prediction performance when compared with tra ditional image-based models. The proposed approach can support dermatologists in clinical decision-making and may also facilitate applications in tele-dermatol ogy and personalized disease monitoring.
Authors - Vijayanirmala Baddala, Jolakula Asoka Smitha, Bichagal Shadaksharappa Abstract - Accurate State-of-Charge (SoC) estimation is critical for ensuring the reliability, safety, and operational efficiency of lithium-ion batteries in electric vehicles and energy storage systems. While data-driven models offer high precision, centralized approaches are increasingly limited by data privacy concerns, high communi- cation overhead, and poor scalability. This paper addresses these challenges by proposing a comprehensive deep learning and federated learning (FL) frame- work for decentralized SoC prediction using the OSF battery dataset. We use four LSTM architectures: Stacked LSTM, Bidirectional LSTM, Attention-based LSTM, and Stateful LSTM, which are integrated into a federated model to sys- tematically evaluate their performance. These include FedAvg, FedProx, and adaptive methods such as FedAdam and FedYogi. To our knowledge, this is the first study to evaluate these architectures in the context of a federated battery management system (BMS). Results show that The comparative analysis inves- tigates the interplay between model complexity and federated optimization, with a specific focus on predictive accuracy, convergence behavior, and robustness to non-IID data distributions stemming from heterogeneous battery capacities and usage patterns. By benchmarking these combinations, this research identifies optimal strategies for implementing privacy-preserving, communication-efficient, and scalable Battery Management Systems (BMS) at the edge.
Authors - Abir Paul, Priti Giri, Rajdeep Ghatak, Soumitra Sasmal, Mauparna Nandan, Partho Mallick Abstract - Accurate forecasting of drug demand is one of the challenging areas in the healthcare service to reduce waste as well as shortages. Some recent studies focused only on predicting drug use demand for regions and hospitals, missing an overall way to combine these forecasts. In this study, a multilevel machine learning framework is presented that merges regional tender demand predictions with monthly and seasonal order forecasting in hospitals and pharmacies. With historical drug usage, the system captures time-based changes, seasonal demands, and also location specific behaviors . Models for regional tenders predict yearly procurement, but models at hospitals and pharmacies try to tell the need of each month, allowing better resource distribution.The rigorous experimental process showed better estimates and forecasting with less error than just making a single-level prediction. This framework helps to make better purchasing decisions and ensures a stable drug supply across healthcare systems. Health departments, hospital chains, and pharmacy groups can benefit from using a model .
Authors - Gagandeep Malhotra, Dharm Singh Jat Abstract - Modern Electronic Warfare (EW) environments are very dynamic, crowded, and hostile, which makes static or centralised spectrum-allocation strategies useless. To tackle these issues, this paper introduces a completely adaptable Federated Deep Q-Network (A-FDQN) framework for each node, which is built onto a high-fidelity ns-3.40 EW simulation environment. In this simulation each tactical radio has been configured to work as an independent federated client which trains a local DQN within itself based on metrics obtained from SINR, congestion, delay, jitter, and interference caused by jamming. A federated server then periodically collects client models using Federated Averaging (FedAvg) or Median method. This lets global learning happen without needing centralised state visibility or constant connectivity, which is very important for networks on contested battlefields. Our framework is different from earlier RL and FL studies because it combines per-node reinforcement learning, periodic federated aggregation, partial client participation, tuneable synchronisation frequency, and realistic ns-3 modelling of mobility, sweep jamming, bursty traffic, congestion hotspots, and Wi-Fi PHY/MAC interactions. Our A-FDQN system dynamically changes the channel assignments at each node, which greatly improves the packet delivery ratio, minimum SINR, fairness, and delay when faced with challenging EW scenarios. This first of its kind end-to-end FRL architecture offers a decentralised, resilient, and operationally relevant way to manage the spectrum for next-generation military communications.
Authors - Harsh Vardhan, Harsh Vikramaditya, Doyelshree Bhui, Shilpi Basak, Soumitra Sasmal, Subhajit Bhowmick, Ishan Ghosh Abstract - Security audits present a unique and ever evolving challenge due to the dynamic nature of cyberthreats and complex regulations. Traditional compliance audits remain largely manual and labor inten sive, resulting in vast inconsistencies. This paper introduces a solution to make compliance audits easier and faster by proposing a framework that leverages the use of Natural Language Processing and Large Lan guage Models to map organizational policies to frameworks and allows for real-time data from security controls to be validated against these complex security frameworks. Through a hybrid multi-model architec ture, the solutions in this paper aim to enhance the accuracy and trans parency of compliance evaluations coupled with evidence-backed insights. The results demonstrate the potential of integrating intelligent auditing systems to deliver compliance assessments that are consistent, accurate, and rapid; streamlining governance and improving cyber security posture management.
Authors - Anjali Yawatkar, Hemlata Gaikwad Abstract - Contemporary customer support systems require processing a massive number of user queries with low latency and high semantic relevance. Rule-based systems fail to capture context, while fully LLM-based systems are computation ally expensive and suffer from high latency. This paper introduces an adaptive AI-assisted customer support automation system using an optimized Retrieval Augmented Generation (RAG) model. The proposed system combines Azure OpenAI embeddings, FAISS-based vector search, selective Cross-Encoder re ranking, and a Learning-to-Rank (LambdaMART) model for adaptive score fu sion. Unlike vanilla RAG models, the proposed system adaptively re-ranks only the top-k retrieved candidates, trading off ranking precision and latency. Experi ments were carried out on a 1,30,000-sample e-commerce customer support da taset with query-response pairs annotated with intent labels. Compared to rule based retrieval, embedding+FAISS, and vanilla RAG models, the proposed hy brid system showed improved top-1 retrieval precision with a concurrent reduc tion in end-to-end latency from 0.414s to 0.365s (≈11.8% relative improvement). The LambdaMART model adaptively learned weights from FAISS and Cross Encoder scores, improving ranking robustness and eliminating misranked top re sponses. The system was implemented on Azure Machine Learning with a cloud scale pipeline and interactive Streamlit web interface, showcasing the cost-effec tive inference capabilities of the proposed system via selective re-ranking.
Authors - Jaykumar Gandharva, Hardika Menghani, Tilak Brahmbhatt, Nischay Agrawal Abstract - Modern Electronic Warfare (EW) environments are very dynamic, crowded, and hostile, which makes static or centralised spectrum-allocation strategies useless. To tackle these issues, this paper introduces a completely adaptable Federated Deep Q-Network (A-FDQN) framework for each node, which is built onto a high-fidelity ns-3.40 EW simulation environment. In this simulation each tactical radio has been configured to work as an independent federated client which trains a local DQN within itself based on metrics obtained from SINR, congestion, delay, jitter, and interference caused by jamming. A federated server then periodically collects client models using Federated Averaging (FedAvg) or Median method. This lets global learning happen without needing centralised state visibility or constant connectivity, which is very important for networks on contested battlefields. Our framework is different from earlier RL and FL studies because it combines per-node reinforcement learning, periodic federated aggregation, partial client participation, tuneable synchronisation frequency, and realistic ns-3 modelling of mobility, sweep jamming, bursty traffic, congestion hotspots, and Wi-Fi PHY/MAC interactions. Our A-FDQN system dynamically changes the channel assignments at each node, which greatly improves the packet delivery ratio, minimum SINR, fairness, and delay when faced with challenging EW scenarios. This first of its kind end-to-end FRL architecture offers a decentralised, resilient, and operationally relevant way to manage the spectrum for next-generation military communications.
Authors - Sachin Kumar Abstract - E-commerce search engines rely on Query Expansion (QE) to bridge the semantic gap between user queries and product catalogs, but expansion can induce query drift, where retrieved results diverge from the user’s original intent. Evaluating QE on novel or out-of-distribution queries is fundamentally intractable under the standard Cranfield paradigm, which requires pre-compiled relevance judgments. This paper introduces the Generalized Authority-Hub Score (GAHS), an unsupervised evaluation metric that repurposes the product catalog’s relational structure— modeled as a product graph—as a dynamic proxy for retrieval quality. Drawing on the HITS algorithm, GAHS quantifies the topical coherence of a retrieved product set without requiring explicit relevance judgments. Using the Amazon ESCI dataset, we validate GAHS against MAP and nDCG@10 on a held-out seen query set, demonstrating strong rank-order agreement (Kendall’s τ = 1.0 with MAP, τ = 0.67 with nDCG@10). We further demonstrate its discriminative power on a disjoint unseen query set, and discuss an observed performance reversal between the two query sets and its implications for QE evaluation methodology.
Authors - K Devi Priya, P Saranya Durga, Y Sony, D Varun Sai Abstract - This paper presents a comprehensive implementation and evaluation of a secure electronic voting system built on the Ethereum blockchain platform. Proposing on Ethereum smart contracts, Proof of Stake consensus, and modern Web3 technologies and implemented the project. The implementation deals with key e-voting issues like voter authentication, ballot privacy, vote immutability and transparent auditability.We examine security threats, offer Layer2 scaling design, introduce concepts of zero-knowledge proofs in order to achieve higher privacy levels, and measure the economic benefit of deployment on different scales. In our results, we have shown that Ethereum has a significant basis to support decentralized voting systems, but scalability and cost reduction remain an important challenge to large-scale elections. The paper ends with a set of practical recommendations on the deployment of production and the main directions on the further research in the field of blockchain-based democratic systems.
Authors - Manav Thakar, Nischay Agrawal, Jaykumar Gandharva, Manish Singh Abstract - Predicting and understanding the inhibitory activity associated with Breast Cancer resistance protein can assist in the drug discovery process by anticipating the potential drug resistance and drug-drug interactions. Prediction of BCRP inhibitors using machine learning can accelerate the identification of BCRP inhibitors by analyzing large datasets, finding patterns in molecular structures, and predicting interactions that would be time-consuming and expensive through traditional methods like high-throughput screening or trial-and-error experimentation. In the literature, machine learning has been employed to develop techniques for predicting BCRP inhibition. However, these methods often exhibit low prediction accuracy, highlighting the need for improved prediction techniques with enhanced accuracy. In this research, BCRP inhibition prediction has been carried out using features spaces fusion to enhance the features information with richer representation of data incorporating complementary aspects of molecule to get the increased accuracy for discovery of inhibitors for drugs of breast cancer. The experimental results show that the proposed technique has increased accuracy and precision for the discovery of BCRP inhibitors. The accuracy of the proposed technique is 97% which is higher than the techniques developed in literature. The study demonstrates that enhancing the features information by combining various compound properties creates a more richer and comprehensive feature space. This enhanced feature representation can significantly help in identifying BCRP inhibitors specifically and contribute to advancements in drug discovery overall.
Authors - Shaik Sohail Ahammed, B. Rohan Teja, R. Naga Sumithra, D. Manasa, T. N. V. D. Sai Krishna Abstract - Privilege Escalation is a major issue for securing Linux sys tems. When a user gains unauthorized root access he has the ability to access all system resources and manipulate them at will. In the past, Linux has used Static Access Control Policies and User Space Monitoring Tools to secure system access. However, these methods provide little in sight into how the kernel is modifying users credentials when permissions are changed. In this paper we propose a Kernel-Level solution to detect and prevent unauthorized privilege escalations. This detection/ preven tion occurs in real time via a Credential Transition Monitoring Mecha nism within the kernel layer, which prevents the elevation of privileges by illegal means. To create the functionality necessary for the above, a Linux Kernel Module (LKM) was created which utilizes kprobes to in tercept calls to the commit creds() function, which is used to update a processes credentials in the kernel. To evaluate if the privilege escalation being requested is legitimate or malicious, the LKM contains a Policy Based Evaluation Mechanism which evaluates each request to modify a process’s credentials. We tested our proposed solution using a con trolled test environment composed of a Virtual Machine (VM) running the Ubuntu Operating System. We ran two types of tests, first were Le gitimate Administrative Operations utilizing the ”sudo” utility, second were Simulated Privilege Escalation Attacks based upon SetUID Vul nerabilities. Our results show that the system effectively detected and blocked malicious privilege escalations, while providing minimal over head to normal system operation.
Authors - Menna Elgabry, Ali Hamdi Abstract - Mortality prediction for intensive care unit (ICU) patients with alcohol-related disorders remains insufficiently explored despite the distinct clinical characteristics and elevated risk profile of this population. Unlike general ICU cohorts, these patients often present with impaired physiological function, frequent complications, and poorer overall outcomes. However, few research works have taken this patient group into account for mortality prediction. This study addresses the gap by developing mortality prediction models specifically for ICU patients with alcohol-related disorders using multimodal electronic health record data. To capture the complex clinical status of patients, we integrate six major data modalities in the first 24 hours after admission, including demographics, diagnoses, medications, procedures, laboratory results/vital signs, and patient outputs. A refined preprocessing pipeline was used to harmonize and process heterogeneous input data. In addition, severe class imbalance is another challenging issue in resolving this mortality predict task. Therefore, our work examines systematically several rebalancing strategies: no resampling, oversampling, undersampling, and SMOTENC. Evaluated on both MIMIC-III and MIMIC-IV databases, our proposed rebalanced multimodal data approach is effective for tackling the task. Indeed, the experimental results show that CatBoost with random undersampling provides the most consistent and balanced effectiveness. Furthermore, multimodal analysis demonstrates that combining diagnoses, laboratory results/vital signs, and medications substantially improves prediction, while integrating all modalities achieves the best overall performance.
Authors - Chinmayee Padhy, Himansu Mohan Padhy, Pranati Mishra, Nabin Kumar Nag Abstract - Establishing an institution's excellence requires measuring their innovation and research accomplishments. Tracking, verifying, and evaluating innovation and research output in an efficient manner is currently constrained by a lack of efficient reporting systems and disorganized methods of obtaining the necessary data. The creation of InnovateHub, a web-based, secure, scalable, and cloud-based platform that provides a centralized system for analysing, managing, and visualizing research and innovation throughout the world's education sector. The InnovateHub provides a central location where a single point of access can be used to collect and process all types of innovation and research information via an effective system; an interactive dashboard and analytical visualisation allows users easy access to relevant information. InnovateHub provides a role and permissions-based access control mechanism to preserve the data privacy and accountability of Administrators, Faculty, and Students. InnovateHub also supports Multi Factor Authentication (MFA) using JSON Web Tokens (JWT) for multiple layers of security and verification of user identity as well as One Time Passcode (OTP) confirmed through email, and uses cryptographic hashing to provide a form of security for storing documents and provides a biometric face-based verification system (i.e., facial recognition) to authenticate a user during critical submission phases. Automated certificate generation and contribution recognition mechanisms at InnovateHub provide additional visibility into, and motivation for, users' contributions to the platform. Utilizing the MERN Stack and AWS for Hosting of MERN Stack: Utilizing the MERN Stack (MongoDB, Express, React, Node.js) & AWS to Host a MERN Stack Application Innovative Hosting Solutions by AWS Include Amazon EC2 Instances to Host Both the Application Back End as Well as Application Front End Services and Amazon S3 for Secure and Scalable Storage of Research Document & Certificate Generation. Experimental Deployment Indicates Reliable Operation, High Availability and Secure Handling of Data During Real Time Utilization within the Loss Prevention Environment. Innovate Hub Provides Real Time Analytics, Secure Verification & Cloud Scaleability for Institutional Research Governance and the Development of a Data Driven Platform of Continuous Innovation and Growth through the Development of a Data Driven Innovation Platform.
Authors - Pranav Rao, Pranav S Acharya, Rishika Nayana Naarayan, Shreya M Hegde, Pavan A C Abstract - The rapid expansion of cloud computing, Internet of Things (IoT), 5G networks, and distributed enterprise infrastructures has significantly in creased the complexity and attack surface of modern networks. Traditional net work security mechanisms—primarily based on static rules and signature-based detection—are increasingly ineffective against advanced persistent threats (APTs), zero-day exploits, polymorphic malware, and encrypted attack chan nels. Artificial Intelligence (AI) and Machine Learning (ML) have emerged as transformative technologies capable of enabling adaptive, predictive, and au tonomous cybersecurity systems. This paper presents a comprehensive technical framework for AI-driven network security. We propose a hybrid architecture in tegrating supervised classification, unsupervised anomaly detection, and deep learning-based behavioral modeling. Mathematical formulations for intrusion detection, anomaly detection, and adversarial robustness are provided. The framework is evaluated using benchmark intrusion detection datasets, and per formance is analyzed using standard metrics including accuracy, precision, re call, F1-score, and ROC-AUC. Results demonstrate that AI-driven models sig nificantly outperform traditional signature-based approaches in detecting zero day and evasive attacks. The paper concludes by discussing adversarial machine learning risks and future directions toward autonomous and self-healing net work security ecosystems.
Authors - Rosa Cristina Pesantez, Estevan Gomez-Torres, Cesar Adrian Guayasamin Abstract - The vast implementation of cloud computing has uplifted the modern IT practices by improving scalability, flexibility, and budget efficiency. In contrast, there has been an increase in energy consumption, which results in carbon emissions. This happens because of overusage, overconsumption, overprovisioning, unused capacity, and inefficient data center management. These days, data centers act as the sole contributor to global greenhouse gas (GHG) emissions; therefore, sustainable cloud operations are essential in addressing this challenge. GreenOps, or green operations, defines the cloud deployment and operational practices that take place but also considers the environmental impact; it depicts energy-efficient infrastructure design, optimized resource usage, virtualization, and the integration of renewable energy resources. This survey presents a summary of green cloud computing, including the current trends, challenges, energy-aware scheduling algorithms, and optimization techniques for obtaining energy-efficient cloud deployment.
Authors - Govind Sambare, Sarika Deokate, Saurabh Dhakite, Sahil Ambokar, Gargi Barve Abstract - Static perimeter-based security architectures are now inef fective in the current threat scenario. The ability of attackers to obtain legitimate credentials and the presence of zero-day exploits often cause real-time breaches of the network perimeter. An area of concern is the real-time monitoring of these systems. In the current scenario, security monitoring is performed in a segregated manner, where network analysts analyze time-stamped network logs and identity analysts analyze time stamped login attempts, without cross-referencing in real time between these two domains. The proposed solution is a fusion platform capable of ingestion of raw network transport data and real-time human element monitoring data. This is achieved through the integration of two dif ferent threat detection mechanisms using a FastAPI backend. The first threat detection system will be the Network Threat Detector (NTD), im plemented in Python and using the Scapy library to parse deep packet data in real time for flow analysis. The second threat detection system will be a JavaScript tracker designed for monitoring digital behavioral indicators and calculating real-time metrics such as mouse velocities, ac celerations, kinematic jerk, and typing speeds. Real-time monitoring will be achieved through a machine learning framework with three different modules for inferring user intent using the Random Forest algorithm, detecting anomalous statistical patterns using the Isolation Forest algo rithm, and detecting malicious plaintext syntax using Logistic Regres sion. The system has been tested in a lab scenario and has been able to classify user session states into four different states: Engaged, Con fused, Frustrated and Suspicious with accuracy exceeding 95%. These digital behavioral indicators will be fed into the Network Transport Data (NTD), allowing the computation of a real-time risk score.
Authors - Duc Thinh Nguyen, Diem Huyen Nguyen Ngoc, Khoa Tran Thi-Minh Abstract - In the present-day context, presentations and computer-based interac tion play a crucial role in various domains, particularly in education and business. Traditionally, users have to rely on physical devices such as mouses, keyboards, or laser. Although these devices meet the basic requirements, they still reveal many limitations regarding mobility, continuity, and dependence on battery life. To address these limitations, hand gesture-based presentation control systems have emerged as a promising solution due to their intuitive, natural, and engaging interaction style. This paper proposes a touchless system that enables users to control common desktop operations as well as presentations in a natural manner using hand gestures captured via a standard webcam. The proposed system lev erages OpenCV for real-time video acquisition and preprocessing, while Medi aPipe framework is employed for hand tracking and landmark extraction. From the experiments, our system can process in real-time with the accuracy of approx imately 92%. As a result, users can seamlessly control slides, use virtual mouse operations, annotate presentation content, and engage with the audience in a more interactive and natural way without physical contact.
Authors - Deepali Lokare, Pankaj Chandre, Prashant Dhotre Abstract - The rapid expansion of digital services has significantly increased the collection and processing of personal data through online platforms such as e-commerce systems, social media applications, and digital payment services. To regulate the use of personal information, governments worldwide have introduced data protection regulations such as the General Data Protection Regulation (GDPR), the Digital Personal Data Protection Act (DPDPA), and the California Consumer Privacy Act (CCPA). Organizations publish privacy policies to inform users about their data practices; however, these policies are often lengthy, complex, and difficult for users to understand. Consequently, users frequently accept privacy policies without fully reviewing how their personal data is collected, processed, and shared. Recent research has explored automated approaches for privacy policy analysis using artificial intelligence techniques, including machine learning, natural language processing, and large language models. Retrieval-Augmented Generation (RAG) has further enhanced compliance evaluation by linking policy statements with relevant regulatory clauses. Despite these advancements, challenges remain, such as the lack of standardised datasets, limited explainability of AI decisions, dependence on prompt design, and insufficient validation with regulatory experts. This paper discusses future research directions in AI-driven privacy policy compliance analysis and highlights emerging opportunities for improving regulatory compliance assessment, user privacy protection, and transparent privacy governance in digital ecosystems.
Authors - Samiksha M, Sharanya G S, Shrina Anahosur, Surabhi K C, Surabhi Narayan Abstract - Multi-angle image synthesis is highly important when it comes to the generation of 3D scenes. But the current methods are either ex pensive in terms of computational costs or lack photorealism in their outputs. We propose a novel sketch and text based multiview image generation approach that solves the above-mentioned problems by mak ing use of multimodal diffusion models efficiently. Our pipeline utilises DreamShaper v8 for converting the input sketch and text into a pho torealistic 2D image and then passes this 2D image into a fine-tuned Zero123plus model for the final generation of consistent multiview im ages, showing a 43.69% improvement in the overall perceptual quality compared to baseline sketch-to-multiview models. Moreover, our pipeline shows flexibility in scalability by generating anywhere from 6 to 64 consis tent multiview images according to the requirements of the downstream tasks. We demonstrate the success of our pipeline through extensive ex periments conducted using voxel-based grid approaches and Neural Ra diance Fields (NeRF). Our pipeline greatly reduces computational costs, all while maintaining photorealism in the outputs, confirming the poten tial of sketch and text based multimodal conditioning as an intuitive and efficient paradigm for controlled 3D content generation.
Authors - Balasubramanian M, Arasu Prabhu V S, Nalini Subramanian Abstract - Privilege Escalation is a major issue for securing Linux sys tems. When a user gains unauthorized root access he has the ability to access all system resources and manipulate them at will. In the past, Linux has used Static Access Control Policies and User Space Monitoring Tools to secure system access. However, these methods provide little in sight into how the kernel is modifying users credentials when permissions are changed. In this paper we propose a Kernel-Level solution to detect and prevent unauthorized privilege escalations. This detection/ preven tion occurs in real time via a Credential Transition Monitoring Mecha nism within the kernel layer, which prevents the elevation of privileges by illegal means. To create the functionality necessary for the above, a Linux Kernel Module (LKM) was created which utilizes kprobes to in tercept calls to the commit creds() function, which is used to update a processes credentials in the kernel. To evaluate if the privilege escalation being requested is legitimate or malicious, the LKM contains a Policy Based Evaluation Mechanism which evaluates each request to modify a process’s credentials. We tested our proposed solution using a con trolled test environment composed of a Virtual Machine (VM) running the Ubuntu Operating System. We ran two types of tests, first were Le gitimate Administrative Operations utilizing the ”sudo” utility, second were Simulated Privilege Escalation Attacks based upon SetUID Vul nerabilities. Our results show that the system effectively detected and blocked malicious privilege escalations, while providing minimal over head to normal system operation.
Authors - Noel Milliones, Vicente Pitogo, Mark Phil Pacot Abstract - The sensitive information in the healthcare industry along with the increasing phe nomenon of the use of intelligent health-related devices makes it a very difficult task to ensure the privacy of patients as well as carry out precise analysis. The centralized methodology in cur-rent machine learning models requires the exchange of raw information of patients from different healthcare institutions and health related devices to the centralized computer system through the network. However, due to the privacy issues and network traffic issues in this methodology, the proposal proposes the development of a privacy-preserving health analytics platform. Here in this proposed methodology, every healthcare center as well as health-related device has its own local machine learning model without transferring even a single piece of information outside. However, the models also employ disease-specific models including CNN heart diseases models of 95 percent accuracy, Gradient Boosting Classifier Diabetes models of 93 percent accuracy models, along with SVM models of liver diseases along with 96 percent GridSearch models. Each edge device carries out the data preprocessing for the local environment, as well as the processes of model training and the transmission of secure updates, in such a way that the sensitive patient data has never left the environment. The platform presented proves the idea that edge computing and collaborative learning can lead to scalable and secure healthcare analytics with high predictive performance.
Authors - Etambuyu Akufuna, Mayumbo Nyirenda, Ruth Wahila, Marjorie kabinga Makukula Abstract - As the primary cause of death worldwide, cardiovascular disease (CVD) necessitates accurate early detection methods. We provide a machine learning approach for predicting heart illness using clinical health data that is enabled by the Internet of Things. An SVM classifier that was trained using 14 Cleveland Heart The disease dataset separates patients at high risk from those in good health. Preprocessing, feature standardisation, and GridSearch Cross-Validation hyperparameter optimisation are all included in the workflow. The model outperforms a number of benchmark techniques in the literature with an accuracy of 93.33% and an AUC of 0.97. A scalable and comprehensible basis for IoT-based clinical decision assistance is confirmed by comparative outcomes.
Authors - Hemamalini Siranjeevi, Swaminathan Venkatraman, Dharshini V, Gayathri A, Sushma Sri R Abstract - Urban environments generate massive video data from surveillance and mobile sensors, necessitating efficient and intelligent summarization for smart city and transportation systems. This paper proposes a multimodal video summarization framework that moves beyond object-centric analysis toward high-level urban scene understanding. Unlike traditional methods that rely on low-level visual features or isolated object detection, the proposed approach captures contextual relationships and temporal continuity through a multi-stage pipeline. The system integrates multimodal perception, combining deep learning-based object detection, multi-object tracking, and acoustic analysis to preserve entity identities and environmental context. We employ relational inference and motion heuristics to model spatial and semantic interactions, which are then structured into a Dynamic Knowledge Graph (DKG) representing entities, interactions, and temporal events. A semantic synthesis module, powered by a transformer-based language model, generates concise, coherent, and semantically meaningful summaries. This architecture enables scalable, context-aware video summarization adaptable to real-world urban applications.
Authors - Nithin Gattappagari, Lakshmi Sagar S, Reddy Lokesh K, Banu Prakash N, Asritha A, Varalakshmi U, Karthik P, Praveen Kumar Rayani Abstract - Conventional one-time authentication cannot prevent session hijacking after login. This paper proposes a session-level impostor de tection framework based on Siamese learning over mouse dynamics for continuous authentication. The model combines statistical behavioral de scriptors with lightweight temporal modeling (Conv1D+GRU) to learn compact embeddings for open-set verification. It supports one-shot en rollment by comparing a query session against a single verified reference session and stores non-reversible embeddings instead of raw trajectories to improve privacy. We evaluate on Balabit and SAPiMouse under se vere class imbalance using balanced batching, semi-hard negative mining, and focal contrastive loss. The framework achieves AUROC 0.95/0.96, F1 0.80/0.85, and accuracy 0.92/0.93, with 46K trainable parameters and approximately 15ms inference time, indicating practical deployment potential.
Authors - Rishav Kumar Agrawal, Maharshi Bhowmick, Mir Abbas Hussain, Sachin, Vaishali Shinde Abstract - This paper presents a platform for scalable validation, visu alization, and explanation of synthetic tabular data in a rigorous and operationally practical workflow. The system integrates statistical test ing, dimensionality reduction, anomaly detection, and AI-assisted in terpretation into a single analysis pipeline. Through an insurance-data case study, we show that the platform can detect subtle distributional artifacts, support utility–privacy trade-off assessment, and provide in terpretable evidence that is difficult to obtain from isolated univariate checks. We conclude by discussing practical value, current limitations, and directions for future development.
Authors - Rowena Ocier Sibayan, Hazel C. Tagalog, Ronald S. Cordova Abstract - As digital marketing expands in Oman, many organizations struggle to transform large volumes of customer data into actionable insights. This study presents an AI-driven marketing intelligence framework designed for non-technical users, combining automated customer segmentation, sentiment analysis, and personalized recommendations. The framework employs an autoencoder-based feature extraction approach to capture key behavioral patterns, followed by K-Means clustering to define meaningful customer segments (Berahmand et al., 2024). A fine-tuned BERT model analyzes multilingual feedback in Arabic and English to assess customer sentiment (Manias et al., 2023). The framework was evaluated using 12 months of campaign data from 450 customers across multiple Omani businesses. Analysis revealed four distinct customer groups and an overall positive sentiment of +0.55. Controlled A/B experiments demonstrated that AI-guided campaigns outperformed traditional methods, increasing conversion rates by 27%, improving retention by 15%, and generating a threefold return on marketing spend. These results indicate that accessible AI tools can deliver measurable marketing benefits in emerging markets and provide a scalable solution for Gulf-region businesses.
Authors - Maria George Anthraper, Kusuma Sanjaykumar, Sinchana K C, V R, Badri Prasad Abstract - Post-quantum migration is increasingly constrained by time: deployed cryptographic mechanisms may need to be retired, hybridized, or re-keyed before effective security margins fall below asset-specific pol icy thresholds. This timing problem is complicated by uncertainty in clas sical hardware acceleration, algorithmic progress, implementation ero sion, and the arrival of cryptographically relevant quantum comput ers. This paper presents a compact probabilistic pipeline that translates evolving assumptions and evidence into decision-facing migration guid ance. The approach couples three layers: (i) a security-trajectory model that encodes expected margin erosion under scenario parameters, (ii) a latent-regime model that represents partially observed risk states and updates them as evidence changes, and (iii) an option-style timing layer that quantifies the diminishing value of delaying migration as thresholds approach. Outputs are conditional on stated assumptions and are in tended to be reported with sensitivity bands and lead-time constraints. In practice, the pipeline is intended to be re-run as assumptions and evidence evolve, preserving an auditable trail from scenario inputs to in termediate states and final decision artifacts. The primary deliverables are comparative rankings and conservative “start-by” windows under stated assumptions, rather than single predicted break dates.
Authors - Jayalakshmi D, N. Priya Abstract - Online product reviews play a key role in the success or failure of an e-commerce business. Often, online reviews from previous customers provide buyers with detailed advice about the product and help them decide before purchasing a product or service. However, some e-commerce products can be promoted or damaged by fraudsters who post fake reviews. Synthetic Reviews (SRs) have the capacity to deceive consumers, influence purchasing decisions, and lead to losses. Thus, SRs pose a significant risk to e-commerce companies and content creators, undermining consumer loyalty and brand reputation. Specifically, the development of AI-generated fake reviews has made them harder to detect, as they are very similar to human-written texts. This review paper presents a Deep Learning (DL)-based framework that offers comprehensive insight into fraud and synthetic review detection in an evolving e-commerce environment. This review paper discusses the importance of DL for detecting online product fake reviews in sentiment analysis using various approaches based on Graph Convolutional Network (GCN), Hierarchical Graph Attention Network (HGAN) Sentiment Majority Voting Classifiers (SMVC), Convolutional Neural Networks with Bidirectional Long Short-Term Memory Networks (CNN-Bi-LSTMs), and a proposed Optimized Bidirectional Encoder Representation Transformers (OBERT) model. This review paper focused on the importance of DL models, particularly the GCN, for effective identification of fake online reviews. This review paper proposed a DL algorithm for fake review detection in online products and demonstrated its practical application in a real-world scenario.
Authors - Miroslav Cech, Rastislav Roka Abstract - Private 5G networks require a reliable, high-capacity, and secure transport infrastructure, especially in industrial and critical applications. Free Space Optics is a promising solution enabling multi-gigabit transmissions with low latency and increased physical security. The article analyses the possibili ties of integrating FSO technology into Standalone Non-Public Network and Public Network Integrated Non-Public Network architectures and evaluates the role of FSO links as a transport or interconnection layer and their impact on la tency, reliability, and security for 5G services such as eMBB, URLLC, and mMTC. The article then summarizes current research trends, including the use of artificial intelligence and machine learning to optimize FSO-based transmission.
Authors - Tanmoy De, Vimal Kumar, Pratima Verma Abstract - The process of operating modern engineering companies is often compartmentalized due to the straightforward nature of the operations requirements that mani-fest themselves within the realm of the software creation and hardware manufacturing. The absence of integration between Agile practices and Waterfall lifecycles is a waste of administrative resources and delays time-to-market. A hybrid project management SaaS is offered in this project called Converge, which will target the integration of these areas without sacrificing the integrity of the data stored in digital code repositories and physical Bill of Materials (BoM). The adoption of Multi-Modal Documentation, Real-time State Synchronization and IoT-oriented Task Automation have their measures of efficiency of workflow, responsiveness of interface, and cross-domain data consistency. The most recent breakthroughs in Natural Language Processing (NLP) and Computer Vision are used to make the experience more practical; a custom AI pipeline based on the ResNet50 and LSTM networks are able to extract visual storyboards of technical video reports with an impressive F Score of 83.00% (with 79.20% Precision and 86.50% Recall), and Transformer based models (including BART) are able to generate structured textual summaries with the leading ROUGE-L score of 0.42. The system is anchored on a dynamic split-brain architecture to display coherent information in either Kanban boards or Gantt charts as the case arises. Status updates increase exponentially with integrated IoT triggers to computerize the execution of tasks via a direct hardware to software communication. The survey is based on the trade offs between the flexibility of UI, the complexity of the database schema, and the latency of the API to compare the old siloed tools to this new hybrid framework. The future of engineering management relies on new tendencies, such as Hybrid Machine Learning, to predictively allocate resources, cutting the error rates in estimating the effort by three times (MMRE to 0.32) with the help of such dominant historical measures of resources as Lines of Code (feature importance score of 0.73) and automated reporting of resource depend-ency. Finally, it is demonstrated that the suggested architecture with the support of a CNN optimized backend video storage, which will save 61.80% of the time at a small cost of 2.30% BDBR, will save about 60% of time on manual docu-mentation and synchronize assets in real-time with a latency less than 200ms (2 seconds).
Authors - Dennis A. Dizon, Gleen A. Dalaorao Abstract - Access to formal financial services remains limited in many develop ing regions, largely due to economic and infrastructural constraints. This study uses the ISO/IEC 25010 as the evaluation framework to present a software quality assessment of a lending automation system installed in a financial insti tution in Butuan City, Philippines. The evaluation focuses on five essential as pects of software quality: usability, reliability, functional suitability, perfor mance efficiency, and security. Usability surveys using SUS and UMUX-Lite, operational and performance testing, and an evaluation of security and data privacy compliance were used to gather empirical data. According to the results, the system achieved high performance with an average inference latency of 0.208 ms per record, uptime reliability of ≥99.5%, excellent usability with a mean SUS score of 82.5, and full compliance with data privacy regulations. Predictive analytics, specifically the Random Forest model with isotonic calibration, further enhanced the automated loan assessment’s interpretability and reliability. The system proved that it is appropriate for real-world applications and can encourage financial inclusion in resource-constrained environments, as it exceeded the intended benchmarks for each quality model. To guarantee the long-term adoption of lending automation technologies, the study emphasizes the significance of thorough software quality evaluation in addition to predictive accuracy.
Authors - Nita Dimble, Satish Narayanrav Gujar Abstract - The fabrication of components across various industries is accom plished through welding. Although welding has been practiced for more than a hundred years, defects may still occur during the welding process. Thus, indus trial standards require welded joints to be inspected and evaluated to ensure their quality and reliability. Conventional ultrasonic testing (UT) has long been widely used in industry for detecting and evaluating defects in weld specimens. Over the last few decades, advances in sensor technology and signal analysis techniques have significantly advanced ultrasonic testing methods. Advanced methods, such as Time Of Flight Diffraction (TOFD), are more likely to detect linear defects. However, one of the major challenges in applying TOFD to the inspection of austenitic stainless steel (ASS) weldments is noise in the signals. Various signal processing approaches have been developed to suppress such noise, each with its own advantages and limitations. In this work, the focus is placed on the applica tion of multi-level discrete wavelet transform (DWT) decompositions with ‘n’- order wavelet filters for de-noising ultrasonic TOFD A-scan signals. The results show that this approach achieves greater improvement in signal-to-noise ratio (SNR) while requiring less computational time.
Authors - Sreenath M. V., Abhigna Suresh Babu, Addanki Naga Sai Greeshmitha, C. R. Ananya, Lakshmi M., Mohan S. G. Abstract - Conventional recipe formats interrupt cooking workflows by requiring repeated attention shifts to external devices. This paper presents Beyond the Cookbook, a Mixed Reality (MR) cooking assistant developed for Meta Quest headsets. The system delivers spatially anchored, context-aware instructions using persistent holographic overlays, synchronized narration, and multimodal interaction including voice commands, controller input, and hand-tracking gestures. By integrating passthrough MR and spatial mapping, the assistant enables hands-free and hygienic guidance directly within the user’s kitchen environment. A usability study with twenty-one participants demonstrates high interaction reliability, instructional clarity, and user confidence. The results validate the feasibility of MR-based procedural learning support in domestic settings.
Authors - Dinesh O. Shirsath, Swati V.Sankpal Abstract - This paper presents a hybrid denoising pipeline for multi-channel electrocardiogram (ECG) recordings. First, blind source separation (BSS) isolates putative sources (cardiac, motion, muscle, baseline drift). Second, each separated component is represented sparsely in a suitable transform or learned dictionary; small / noise-dominated coefficients are attenuated and the component reconstructed. Finally, recombination yields a denoised ECG that preserves waveform morphology while suppressing compound, nonstationary noise. The paper describes the mathematical model, algorithmic steps, implementation tips, evaluation metrics, and practical considerations for deployment.
Authors - Aarya Sagar Sonawane, Rutuja Rajendra Thorwat, Shravani Rajeev Deshpande, A. R. Bankar Abstract - A significant security issue facing organizations is insider threats since one has access to privileged information and the behavior of users keeps evolving. Current solutions can be un-explainable, unable to manage new behavior patterns, generate high false positives, and un privacy friendly because of centralized data analysis. To solve these problems, this paper presents EXPLAIN-ITD, an explainable, adaptive and privacy-aware artificial intelligence system to detect insider threats. The framework is an integration of multi-modal data fusion, dual memory continuous learning, explainable risk scoring, human feedback in the loop and federated learning and differential privacy. As the exper imental findings have demonstrated, EXPLAIN-ITD has a better level of accuracy in detection, a lower level of false alarms and better interpreta bility than the current approaches.
Authors - Kamalakar S, Anjan Babu G, Ravi Kumar G Abstract - Artificial intelligence has become an important tool for addressing environmental challenges because it can analyze large datasets, detect patterns, and support accurate predictions. As climate change increases pressure on natural and built environments, organizations adopt AI to improve monitoring, optimize resource use, and inform sustainability decisions, though research remains fragmented. This review examines studies from 2020 to 2025 and assesses how AI is applied in renewable energy, water management, agriculture, waste management and the circular economy, and environmental health and public safety. A major objective of this synthesis is to highlight commonly employed functions by researchers and practitioners such as forecasting, anomaly detection, and operational optimization, alongside emerging model frameworks that strengthen environmental management. While AI offers meaningful benefits, it also presents challenges related to governance, transparency, and the energy demands of large scale models. This review consolidates developments and identifies priorities for future research.
Authors - Anil Kumar Bandani, Anupama Bollampally, Ramesh Deshpande B Saritha, P Rajesh Abstract - Transformer-based models in modern applications struggle with continual learning due to catastrophic forgetting. This paper presents Lapis Whale, a framework that incorporates a Selective Replay Utilization Mechanism (SERUM) to help a model retain previously learned knowledge while adapting to new tasks. The approach leverages a memory buffer to replay representative samples from earlier tasks during training. Experiments on the CIFAR-100 dataset show improved accuracy retention and reduced forgetting compared to standard fine-tuning methods. The framework is computationally efficient and well-suited for real-world adaptive AI systems.
Authors - Suman Kumar Mandal, Wendrila Biswas, Jaydev Mishra Abstract - Glaucoma is an optic neuropathy that is progressive and one of the most common causes of permanent blindness in the world. The retinal fundus images used to diagnose the condition are still time-consuming and highly reliant on the clinical expertise to detect the condition early, before the loss of vision becomes severe. In this experiment, we suggest a deep learning model that will use the ResNet50 architecture to identify retinal fundus images as belonging to one of two categories: Referable Glaucoma (RG) and Non-Referable Glaucoma (NRG). ResNet50 has been selected because it has good feature ex-traction (residual learning and deep convolutional learning). The standard performance measures were used to assess the trained model, such as accuracy, precision, recall, F1-score, and area under the ROC curve. The experimental findings indicate that the suggested approach yields consistent and accurate classification of RG and NRG cases, and it can be used to assist the ophthalmologist in clinical decision-making. The paper demonstrates how deep learning models could assist in further development of early glaucoma detection and mass screening, which, in their turn, can contribute to better patient outcomes and prevention of blindness before its onset.
Authors - S. Jayaraj, G. Anjan Babu, Krishnamurthy Kavitha Abstract - As neurodegenerative diseases like Huntington’s become a global health priority, the difficulty of early and accurate radiological diagnosis remains a significant hurdle. While Deep Learning, predominantly CNNs (Convolutional Neural Networks), offers a clarification for medical image classification, performance is often hindered by the inadequacy of high-grade datasets. This research addresses these limitations by proposing an ensemble deep learning model that integrates ResNet, MobileNet, and VGG16 architectures. By combining these networks, the study achieves enhanced robustness and superior classification accuracy compared to standalone models. This automated framework serves as a vital clinical support tool, enabling faster interventions, improved treatment planning, and a reduction in the global burden of neurodegenerative disorders [10,12].
Authors - Abhijit Dnyaneshwar Jadhav, Prashant G. Ahire, Madhuri Hiwale Abstract - A significant security issue facing organizations is insider threats since one has access to privileged information and the behavior of users keeps evolving. Current solutions can be un-explainable, unable to manage new behavior patterns, generate high false positives, and un privacy friendly because of centralized data analysis. To solve these problems, this paper presents EXPLAIN-ITD, an explainable, adaptive and privacy-aware artificial intelligence system to detect insider threats. The framework is an integration of multi-modal data fusion, dual memory continuous learning, explainable risk scoring, human feedback in the loop and federated learning and differential privacy. As the exper imental findings have demonstrated, EXPLAIN-ITD has a better level of accuracy in detection, a lower level of false alarms and better interpreta bility than the current approaches.
Authors - Tirupathi Rao Dockara, Pradeep Rajagopal Kirthivasan Abstract - Healthcare data scarcity poses significant challenges for machine learning applications in clinical settings, particularly for conditions with limited patient populations. This paper presents a novel quantumenhanced data augmentation framework that addresses this challenge through a three-pillar architecture: Quantum Random Number Generation (QRNG) for true randomness, Statistical AI for intelligent parameter optimization, and Generative AI for clinical interpretability. Our implementation utilizes Bell state quantum circuits to generate genuinely random perturbations, ensuring higher entropy than classical pseudorandom methods. The framework incorporates medical domain knowledge through constraint-aware augmentation, maintaining clinical validity while generating synthetic patient records. Experimental evaluation on the Pima Indians Diabetes dataset (768 samples, 8 features) demonstrates that our quantum-enhanced approach achieves 100% medical constraint compliance while generating high-quality synthetic data. The system provides both command-line and web interfaces, with automatic fallback to classical methods when quantum resources are unavailable. Our contributions include: the first practical application of quantum computing to healthcare data augmentation, an AI-driven optimization system that automatically determines augmentation parameters, integration with large language models for non-technical summarization of validation reports, and a production-ready implementation with comprehensive validation mechanisms. The framework represents a significant advancement in synthetic medical data generation, offering a scalable solution for addressing data scarcity in healthcare AI applications.
Authors - Jyotiprakash Mishra, Sanjay K. Sahay, Swati Mishra, Aman Pathak Abstract - Memory encryption is a key security requirement for modern computing systems, addressing vulnerabilities between CPUs and main memory. Traditional storage encryption is insufficient for protecting volatile data in RAM, which remains exposed to bus sniffing, cold boot attacks, and side-channel exploits. This paper therefore systematically reviews memory encryption techniques focused on hardware-based solutions like Intel Total Memory Encryption (TME), Multi-Key TME, and AMD Secure Memory Encryption, which provide robust protection while minimising performance overhead. The paper also explores integrity protection via Merkle trees and side-channel countermeasures against Differential Power Analysis and Simple Power Analysis attacks. Additionally, granular memory encryption methods for multi-tenant environments are discussed, highlighting their role in isolating sensitive data across security domains. By examining security guarantees and performance trade-offs, we emphasise the necessity of efficient memory encryption to safeguard against evolving threats targeting the CPU-memory interface, providing hardware engineers a foundation for ensuring data confidentiality and integrity.
