Conference Proceedings

Permanent URI for this collection

https://dir.muni.ac.ug/handle/20.500.12260/18

Browse

Now showing 1 - 20 of 25

Improving IoT data security with blockchain and SHA-3-based integrity validation for real-time transmission
(IEEE, 2026-04-07) Dhamodharan, R; Charanya, P; Rajeswari, S V; Guma, Ali; Christy, A. Ananthi; Priyanka, Thella Preethi
The increased exponentially use of Internet of Things (IoT) tools has given serious considerations to the issue of security and integrity of data transferred on real time networks or questions of authenticity of data. This study suggests a secure messaging architecture of combining Blockchains with integrity validation protocol of SHA-3 assuring non-manipulated IoT real-time data streaming. The architecture proposed here would leverage the Keccak based SHA3-256 cryptographic hash algorithm to establish a unique data fingerprint to each transmission that it would then store on a Blockchain ledger through Proof-of-Authority (PoA) consensus mechanism. This enables receivers to match the hashes to validate authenticity of data received on the one hand, and compute a new hash to check the one attached by the sender on the other hand. It was then constructed with Python 3.13 and the libraries used included: hashlib, pandas, matplotlib, and time and simulations were carried out on synthetically produced sensor values representing temperature and humidity measurements of 10 abstract IoT devices. The data packets contain nonce, timestamp in real-time, and unique device ID to increase the entropy. The process of integrity validation and blockchain logging was tested with pace of hash computing, the delay of blockchain logging and the authenticity of verification. It was experimentally observed that 100 percent data integrity was achieved by the system in normal circumstances where hashing time of the system was on the order of 0.11 ms, blockchain logging latency ranged between 50 to 150 ms, and verification was as observed < 0.06 ms. These findings confirm the viability of the combination between blockchain and SHA-3 in lightweight, secure, and real-time communication of IoT. There is a guarantee of reproducibility and scalability to be deployed in real-world systems (home automation in smart homes, industry automation, and patient monitoring).
Influence of organizational culture on employee retention: A Behavioral and data-driven perspective
(IEEE, 2026-03-19) Sumidha, E; Majnoor, Nisbath; Varun, T.; Ali, Guma; Madhavi, N. Bindu; Babu, M.Dinesh
The culture of a company has a big effect on how employees behave, how happy they are with their jobs, and how long they stay with the company. Companies that have a clear culture that puts their employees first usually have more loyal and productive workers and fewer workers who quit. People agree that organizational culture has an effect, but there aren't many real-world studies that use behavioral analytics and predictive modeling together to measure the link between cultural traits and job retention in a quantitative way. Surveys and qualitative evaluations are the main types of methods that are available right now. They can't be made bigger or used to make accurate predictions. This study suggests a mixed analytical framework that uses psychometric surveys and machine learning classification techniques to look at how organizational culture affects employee retention. We used three steps: (1) we collected data with the Organizational Culture Assessment Instrument (OCAI); (2) we used a Random Forest Classifier with employee retention datasets; and (3) we compared the results to those from Logistic Regression, Decision Trees, and Support Vector Machines (SVM). The results showed that the new Random Forest-based model was better at predicting retention than the old ones, with an accuracy rate of 91.3 %. The study found that clan culture, how well people talk to each other, and the way leaders lead all had a big impact on how many employees stayed with the company.
Detecting and mitigating bias in machine learning models
(IEEE, 2026-03-03) Mulekar, Ashwini; Salati, Abid; Tankala, Divya Kumari; Kumar, Gotte Ranjith; Penubaka, Kiran Kumar Reddy; Ali, Guma
The risks of bias in machine learning models are considerable, particularly in sensitive fields such as health care, employment, and finance. Unfair outcomes from these ML/AI systems can exacerbate existing social inequalities. As the goal is to find potential ways to develop a holistic fairness-aware AI framework that can detect, mitigate, and monitor algorithmic bias throughout the ML pipeline, this work proposed a method that integrates causal inference, adversarial debiasing, human-in-the-loop processes to generate feedback, federated learning, and real-time detection of bias drift - all of which can reinforce fairness while minimising the impact on performance. Importantly, the study's experimental results also demonstrated a reduction of similar bias metrics: Disparate Impact decreased by 31%, Equalized Odds Difference decreased by 36%, while delivering an F1-score of your expected 89.1%. To reiterate, this work demonstrated the framework's capacity to create equitable outcomes, with relatively minimal performance sacrifice, across many ML models. The study illustrated that ethical and regulatory issues need to be embedded into the deployment of an AI system and provided a scalable, privacy-preserving framework for organizations to use to build more trustworthy, transparent, and socially responsible ML systems.
SolarSalt: Phase-change sodium-based battery for 24/7 renewable energy storage
(IEEE, 2026-02-20) Roy, Jainish; Obidhusin, Saef; Badriddinovich, Kosimov Khusniddin; Thakur, Diksha; Bhuvaneshwari, K.S.; Ali, Guma
Conventional lithium-ion or high-temperature sodium batteries will not suffice for the growing demand for sustainable and efficient energy storage; research is expanding beyond these. SolarSalt achieves an energy density of ≈450Wh/kg, Coulombic efficiency above 98.5%, and a cycle life exceeding 2000 cycles at room temperature. Compared to lithium-ion batteries, it offers 40% lower kWh costs and improved safety, with no thermal runaway or dendrite formation. By also enhancing stability through HEA coating, which prevents corrosion and dendrite formation to improve the battery lifespan and efficiency, the coating is markedly beneficial. The innovation also offers a low-cost, high-density renewable energy storage alternative to provide uninterrupted power to solar and wind power systems at an appropriate power density. Other areas investigated are legal (such as intellectual property rights, safety certifications (UL, IEC, NFPA), environmental policies (EPA, REACH), and grid interconnection standards (IEEE 1547)). This paper proposes a blockchain-based energy certification system to provide transparency, compliance, and secure energy trading. Experimental results show that SolarSalt outperforms lithium-ion and conventional sodium batteries in terms of energy retention, safety, and cost-effectiveness. This research introduces a novel room-temperature, sodium-based SolarSalt battery that integrates liquid-solid phase change energy storage with HEA encapsulation and blockchain-enabled certification. The originality lies in uniting materials science, renewable energy storage, and digital compliance within a single scalable platform. This contribution advances both technical and regulatory dimensions of sustainable energy storage.
AI-Driven real-time social media analytics for consumer trust enhancement.
(IEEE, 2026-02-20) Sharon; Gnanaroy, E. Rushit; Gupta, Anish; kizi, Sotivoldiyeva Sarvinoz Kahramon; Kannimuthu, S.; Ali, Guma
In a hyperconnected digital ecosystem, companies have never faced as much pressure to maintain consumer trust as they do at present, as content surrounding social media spreads at an alarming pace, along with opinions, feedback, and false information disseminating in real-time. Conventional monitoring approaches, based on batch processing and periodic review, are insufficient for monitoring the dynamism and rapid pace of online interactions. This factor often leads to sluggish reactions to reputational risks and consumer dissatisfaction. In response to this issue, we propose an AI-based, real-time social media analytics system that will continuously track multimodal data streams, including text, photos, and video, to detect signs of consumer mood and trust occurrences. The system combines contextual sentiment analysis using transformer-based natural language processing models, mapping social influence and information propagation through graph neural networks, and anomaly detection algorithms to detect sudden changes in perception or possible misinformation. Streaming pipelines based on distributed computing infrastructure can guarantee lowlatency processing, and real-time alerting and predictive modelling can ensure proactive engagement approaches to reduce risks as they occur. Throughout the simulations, the framework outperforms previous analytics systems, achieving accuracy, precision, recall, and F1-score scores in the 85-90% range across various social media platforms. The system enables organisations to react proactively to negative sentiment, avoid reputational loss, and strengthen consumer trust in real-time by delivering actionable insights and instant alerts. The findings affirm that the combination of developed AI and real-time data processing will provide a scalable technology-based methodology that can not only increase operational efficiency but also provide the company with a sound mechanism for maintaining consumer trust in the dynamic online environment.
AZTEM: A Self-evolving zero-trust model for adaptive cloud security using AI-driven threat mitigation and quantum-resilient encryption
(IEEE, 2026-02-20) Khemraj, Ikhar Avinash; Buckshumiyan, A.; Balassem, Zayd; Shekar, A.; Arunkumar, B.; Ali, Guma; Jabborov, Nurboy
The new digital ecosystems are being based on cloud environments but they are highly vulnerable to the shifts in cyber threat, which can take advantage of the stationary trust models, loose access rules and encryption functions that hangs in the air with the advent of quantum computing. The traditional models of perimeter based security fail to operate in highly dynamic cloud environments with insider based mauling, cross-lateral mauling and the zero day mauling bypassing the traditional security models. In this work, A Self-evolving Zero-Trust Model of Adaptive Cloud security with AI-based Threat Mitigation and Quantumresilient Encryption is introduced and is capable of addressing these emerging problems. AZTEM eliminates the implicit trust, validating and permitting all user, device, and microservice relations through the utilization of a contextsensitive trust engine that is reinforced by reinforcement learning algorithms. To counter the rapid adaptations of attack vectors, the system relies on the deep-learning-based anomaly detection methods such as LSTM-based sequence classifiers on cloud workload telemetry, network flow logs, and publicly accessible datasets (UNSW-NB15 and CICIDS2018). An adaptive policy controller is the basis of dynamic orchestration of mitigation strategies and uses AI-based response mechanisms to quarantine, reroute, or restrict malicious sessions in real time. Besides, the model includes quantum-resilient encryption that extends the protection of both data-at-rest and data-in-transit by using lattice cryptography to secure quantum decryption threats where confidentiality is ensured at the long-term scale. An initial deployment of AWS on Kubernetes clusters demonstrated that threat detection accuracy and false positives decreased by 27 percent and 34 percent respectively over a baseline zero-trust deployment, and performance overhead is less than 8 percent. Its application in adversarial context was assessed through measures such as precision, recall, F1-score and confidence intervals and proved it to be qualified. The architecture is not only making the cloud more resilient to the present, and the infrastructures of the quantum age impervious to the threats of the quantum age, but it is also providing businesses, governments and hard-to-secure areas in need of cloud security without a trade-off an ethical and scaleable and deployment-ready path.
AI-Powered collaborative robotics for next-gen industrial automation
(IEEE, 2026-02-20) Ramya, M; Subburam, S.; Patel, Pushplata; Seetaram, J.; Singh, K. Ranjith; Ali, Guma; Salami, Zayd Ajzan
The runaway development of Industry 4.0 has brought very interconnected, intelligent, and automated production systems in place, but conventional robotics based on rules cannot react to dynamic production conditions, unpredictable working processes, and real-time disruptions, which tend to result in lower efficiency, time delays, and bottlenecks in the work process. Among common challenges faced by factories are inefficiencies in task allocation, the absence of collaborative coordination among robots, poor predictive maintenance, and vulnerabilities in data exchange during interactions. To address those issues, this research proposes AI-Powered Multi-Agent Collaborative Robotic System (AI-MACROS), a highly sophisticated industrial automation system integrating swarm intelligence, reinforcement learning, digital twin simulation, and blockchain-based security. AI-MACROS allows teams of robotic agents to work together, learn dynamically based on environmental interactions, optimize task distribution and deal with unexpected events in the real world. Through digital twins, it provides virtual replicas of robots and work processes, with which to monitor the system in real time, preemptive maintenance and simulate scenarios without disrupting the existing processes. Also, blockchain protocols guarantee integrity and safety of communication and data across the industrial network. Compared to a typical rulebased robotics, one may observe that AI-MACROS is executed by the high performance score (i.e. higher efficiency, accuracy, recall, and F1-scores), and the coordination and energy efficiency is additionally simplified with the help of particle swarm optimization, genetic algorithms and deep reinforcement learning. The proposed solution addresses the disadvantages of the traditional automation, making the operations more resilient in addition to improving the security and scale, not mentioning that it is deemed as not only a potential game changer but also as a next-generation smart factory solution. In a clever synergy, predictive analytics and safe decentralized information processing, AI-MACROS creates high-performance prototype of adaptable, fruitful and credible automation of production under realistic production scenarios.
BERT and CNN for automated detection of detrimental discourse
(IEE, 2026-02-19) Devarapalli, Thejasree; Yarlagadda, Nagalakshmi; Haldorai, Anandakumar; Sivaramakrishnan, A; Ali, Guma; Sharma, Vinod
The modern online age is competitive and full of information that requires sensing dangerous materials in order to maintain online honesty and facilitate positive forms of communication. This paper presents a new model that takes a hybrid approach to using BERT Base embeddings and convolutional neural networks (CNNs) to deal with the problem of detecting harmful material in a variety of settings. BERT Base embeddings are used to dynamically gain the fined tuned features of semantics of text that is then processed by a CNN to classify. Compared to the previous studies mainly concentrated on individual contexts with LSTM, RNN, or more advanced transformer models, we use the effectiveness of CNNs to recognize patterns in contexts and also work with different types of harmful content at the same time. To evaluate it, we applied benchmark datasets a reputable Mendeley dataset of hate speech detection, as well as Kaggle datasets of cyberbullying and emotional distress. The experimental results point to the power of our method, which reached almost 92 percent of accuracy in categories. This study merges semantic expressiveness and CNN effectiveness and contributes to content moderation policies and meaningful insights to create safer and more inclusive websites.
Wearable Nanogenerators power health monitors in off-grid regions
(IEEE, 2025-12-29) Lamba, Akshit; Shamya, A; Fallah, Mohammed H.; Bahodirkhonugli, Sayfiddinov Izzatullakhon; Nallakumar, R.; Ali, Guma
Health monitoring devices in remote areas often don’t have reliable power, making it hard for healthcare staff to help these patients promptly. Because some regions lack reliable electricity, healthcare workers usually struggle to monitor heart rate, blood pressure, and temperature. We suggest combining wearable nanogenerators with health monitoring systems so that the user's motion powers them. Combining triboelectric and piezoelectric principles, the tiny devices can provide the electrical power needed to operate health monitors when people walk or move their arms. This system's algorithm for managing energy permits the devices to operate longer. Simulations of different motions confirm that the proposed system can provide sufficient electricity for the health monitors to run independently. This solution works best for hard-to-reach or underserved areas, providing a more sustainable and affordable alternative to standard power-dependent health devices. The novelty of this study lies in the integrated approach of coupling hybrid piezoelectric– triboelectric nanogenerators with an adaptive energy management algorithm designed specifically for wearable healthcare devices. Unlike prior works that focus primarily on material enhancement or single-source energy harvesting, this research emphasises a co-optimised framework that integrates motion-based energy conversion, storage regulation, and power utilisation control. The contribution of this work is the development of a self-sustaining, algorithm-governed wearable system capable of reliable health data monitoring in off-grid and energy-scarce environments.
Real-time power analytics prevent blackouts in overloaded urban grids
(IEEE, 2025-12-29) Balassem, Zaid Ajzan; Vij, Priya; Kumar, S. Senthil; Rakhmanovich, Ibragimov Ulmas; Pushpalatha, A.; Ali, Guma; Karimova, Farida; Arnav, Jain
Increased demand for energy, factors that change load, and aging infrastructure are putting tremendous pressure on urban power grids. Because of these problems, regular blackouts and trouble keeping the power on occur in large, crowded cities during high-demand times. Typical grid management systems react to events, as they depend on delayed data and mostly manual actions to avoid cascading failures after an overload occurs. To manage these real-time threats, we present in this paper a unique Real-Time Power Analytics Framework (RTPAF) that continuously observes the grid with smart meters and edge computing, uses LSTM neural networks to forecast possible overloads, and sets automatic load redistribution actions using intelligent controllers. A multi-staged framework connects fast data capture, noise reduction before analysis, predictive tools, and critical system response to prioritize hospitals and transport networks. A simulation of an urban grid with 500 nodes, built in GridLAB-D and MATLAB Simulink, was performed to check how the system operated. The simulation found that RTPAF brought down the number of typical blackouts by over 90%, and because its reaction was less than 500 milliseconds, it quickly mitigated possible overload situations. As a result, the model's forecasting accuracy of 94.3% significantly improved the grid’s ability to plan and make decisions. Using this approach in real time strongly supports energy security, minimizes cases where power is interrupted, and can meet the high reliability requirements for future smart cities. The solution suggested is a significant achievement for the preemptive management of urban energy.
Agricultural waste to biofuel: Transforming crop residue into next-generation aviation fuel.
(IEEE, 2025-12-29) Maury, Shyam; Mangaiyarkarasi, V.; Madaminjonugli, Bakhriddinov Makhamadali; MuhamedAle, Hasssan; Arunkumar, E.; Ali, Guma; Rakhimov, Navruzbek; Shetty, Chinmai
The rapid expansion of international aviation has significantly contributed to greenhouse gas emissions. The existing type of jet fuel consumes a considerable amount of crude oil, which has compounded the demand. As a result, the industry accounts for a carbon footprint that reduces the pressure on alternative energy sources with lower carbon emissions. Agricultural waste comprises straw, husks, and stalks, which denotes one such available lignocellulosic feedstock that is not exhaustively utilised to offer a solution to environmental and economic dilemmas in the aviation energy production. The given research paper proposes an integrated solution to transform agricultural waste materials into high-energy-density biofuels, which can be utilised in the aviation industry through a two-stage biochemical and thermochemical treatment, followed by subsequent fuel upgrading to produce a high-quality product. Pre-treatment options, which were attempted on a lab scale, included steam explosion, dilute acid hydrolysis, enzyme saccharification, fermentation, pyrolysis, and Fischer-Tropsch catalytic upgrading. The parameters of the process conditions were optimised to achieve a high yield and minimise energy consumption. Results were statistically analysed to ensure reproducibility, and fuel properties were compared to ASTM D7566 standards to verify that they conformed to conventional jet fuel specifications. The results show that biofuels produced from agricultural waste have an energy density similar to that of Jet A fuel, with notable reductions in carbon and particulate emissions, making them a viable option for mitigating greenhouse gas emissions caused by aviation. The techno-economic analysis also demonstrates the viability of large-scale implementation, based on the availability of feedstock, process effectiveness, and compliance with regulations. Twith regulations. The practice is also compatible with the principles of the circular economy , which emphasises the value of agricultural residues, agrarian eco nomies, and sustainable waste management. Moreover, it is possible to optimise it with AI, emulate the use of blockchains to track feedstock, and adopt the concept of hybrid biofuel electric to make the future of biofuel work more efficiently and easily tr acked. Comprehensively, the paper demonstrates that agricultural waste can be a feasible and sustainable aviation biofeedstock of the next generation, as it can help minimise carbon footprints, make biofuels economically viable, and promote the current trend of carbon neutral aircraft in the global community.
AIoT-driven smart agri-grid (ASAG) for sustainable precision agriculture
(IEEE, 2025-12-29) Sundaram, N. Kalyana; Rajendran, Megala; Ehssan, Muhamed; Soy, Aakansha; Anandhi, K.; Begum, T Ummal Sariba; Ali, Guma; Dhananjaya, B
By advising and teaching farmers on how to apply modern farm practices that embrace Artificial Intelligence (AI) and the Internet of Things (IoT), precision agriculture is revolutionising sustainable farming by optimising for usages that are as much as possible and waste as little as can be afforded. In this research, we propose an AIoT-driven Smart Agri Grid (ASAG) framework that integrates real-time nanosensor networks, an AI-operational control microclimate, an autonomous decision-support system, and secure data sharing via a blockchain using encrypted statistical data. To achieve real-time analytics, edge computing is used in the framework for real-time data analytics, predictive algorithms for dynamic irrigation & nutrient management, and federated learning for distributed AI training, which maintains privacy and scalability. In addition, the system uses AI-based waste-minimisation techniques, such as predictive harvest timing and the conversion of bio-waste into organic fertilisers, thereby reducing post-harvest losses. Experimental results show that ASAG can improve crop yield by 20 to 30%, reduce water waste by up to 50%, and reduce chemical overuse by up to 30%, with its economic and environmental benefits. The feasibility of such deployment on a large scale in precision agriculture is further confirmed by a cost-benefit analysis. The results reinforce the power of AI and IoT in transforming contemporary farming into a self-optimising, climate-resilient system. For long-term sustainability in global agriculture, quantum AI will be used to predict soil health, monitor AI-assisted carbon sequestration, and enable genomic AI for climate-resistant crops.
Augmented heritage: AR and QR code integration for interactive cultural storytelling in the UAE
(IEEE, 2025-12-05) Lal, Mily; Dash, Soumyakant; Keerthika, J; Kumar, Bura Vijay; Ali, Guma; Shukla, Vinod Kumar
This study has outlined a mobile-augmented reality (AR) storytelling framework utilizing Quick Response (QR) code technology to provide cultural engagement at heritage sites in the United Arab Emirates. The project set out to merge traditional heritage with augmented immersive technology and focused on a four-phase methodology: (1) Content curation and narrative design through historians and cultural experts; (2) Planning QR codes in spaces cooperatively designed through spatial design and semiotics of culture; (3) Content development of AR experiences in Unity and Vuforia using finite state machine for stability and usability; and (4) User testing in three major heritage locations. The quantitative and qualitative evaluations measured user engagement, narrative retention, usability of the system, and cultural sensemaking. Relative to baseline digital experiences, the AR-QR platform saw significant gains: 25-30% improvement in engagement; 20-25% improvement in narrative retention; and over 15% improvement in usability and cultural relevance. The study evidence demonstrates the promise of AR-QR storytelling and approach to deliver contextually rich and interactive heritage experiences. Although there are limitations in device affordances/conceivability and longer-term evaluations, this study suggests---immersive storytelling can make heritage more engaging, accessible, and meaningful to a diverse population.
Multi-modal biometric authentication integrating gait and face recognition for mobile security
(IEEE, 2025-10-28) Nivedita, V.; Joseph, Juno Ann; Varghese, Divya Thankom; Nithyanandh, S; Sundaram, Jayavelu; Ali, Guma
In the surroundings of mobile security systems, traditional biometric methods such as fingerprint and facial recognition can be readily disputed and misled. These systems change user appearance by failing in different illumination conditions or spoofing attacks with false fingerprints or images. Overcoming these constraints, the proposed multi-modal biometric identification system integrates gait with facial recognition. Combining facial data with gait traits observed by accelerometers and gyroscopes helps to increase security and reduce false positives and negatives. Results show the system achieves 97.8% accuracy, significantly above the existing systems using a low false acceptance rate (significantly) of 1.5% and a false rejection rate (FRR) of 0.7%. It has a superior resilience to spoofing techniques, including stride imitation and image/video spoofing; it also shines in tough conditions, including poor lighting and appearance alterations. Despite a slight increase in processing time, the method offers a more robust and reliable solution for mobile authentication.
Creating electric vehicle battery management with IoT: Using intelligent algorithms to enhance safety, efficiency, and charging time
(IEEE, 2025-10-28) Balapriya, S.; Pandi, V. Samuthira; Sabitha, M.; Veena, K.; Balasubramaniyan, R.; Ali, Guma
The expanding electric vehicles (EV) market has raised the demand for better-optimized intelligent battery management systems (BMS) that can improve safety, performance, and charging time. Older BMS representative solutions employ earlier monitoring and control techniques, often without real-time adaptability and predictive capabilities. This paper investigated Internet of Things (IoT)-specialized and smart algorithms-integrated EV battery management to increase efficiency, safety, and a superior charging process. The proposed architecture uses machine learning model, data analytics, and IoT-enabled sensors to enable real-time monitoring of critical battery parameters such as state of charge (SoC), state of health (SoH), temperature, and voltage variations. Predictive analytics enable the early detection of potential battery degradation, minimize thermal runaway, and enhance energy distribution among individual battery cells. Furthermore, advanced charging algorithms optimize charging rates in response to instantaneous battery states and grid demand, maximizing charging speed while avoiding overcharging and wearout. Cloud-hosted IoT platforms enable remote monitoring and data-based decision-making, improving user experience and prolonging battery life. We develop a prototype implementation to showcase the effectiveness of the system, which results in efficient energy management, early faults detection, and reduced charging cycles. Application of the proposed system was compared against the state-of-the-art BMS and used as a reference standard and the comparison results revealed excellent performance in terms of safety, compactness and flexibility of the system with the existing BMS. The study utilizing IoT as well as artificial intelligence helps to further emerge smart electric vehicle technology that leads human being for sustainable and rich electric mobility solution.
AI-driven pathogen detection systems for rapid and accurate diagnosis
(IEEE, 2025-09-24) Deepika, M.; Murugan, Sundara Bala; Subasini, V.; Rufus, N. Herald Anantha; Atkinswestley, A.; Ali, Guma
Pathogen detection at speed along with precision serves as a basis for disease diagnosis and overall control effectiveness. Several current detection methods based on culture-based techniques and PCR face drawbacks of extended processing times, high expenses, and the need for expert operators. It adopts deep learning Convolutional Neural Networks with an attention mechanism in a system designed to overcome existing limitations of pathogen detection. The model was trained using a multi-modal dataset that combined spectral and biological signals through feature extraction optimization and privacy-protected training methods based on federated learning. The proposed system reaches a 96.8% detection accuracy together with 15% reduced false alarm rates and 35% faster operation speeds than conventional models. Experimental data measurements showed that this system achieved 92.3% F1-score together with 97.1% AUC for real-time identification of pathogens. AI has proven its effectiveness through these findings in changing the way pathogen diagnostics operate. Further work will concentrate on extending datasets while creating real-time deployment protocols for clinical use.
Next-Gen smart homes: AI-enhanced Li-Fi for superior efficiency and protection
(IEEE, 2025-09-04) Rajasubha, J.; Sathish, A.; Jackson, Beulah; Hemavathi, U.; Ajay, A. P.; Ali, Guma
The system proposed optimizes dynamically the Li-Fi transmission parameters using a Deep Reinforcement Learning (DRL) adaptive modulation algorithm, resulting in a bit error rate of 0.00034 and a throughput of 986.45 Mbps. An intrusion detection system can enhance cybersecurity by utilizing Convolutional Neural Networks, which yield a detection rate of 99.81%. The future paradigm is more secure, reliable, and programmable than conventional Wi-Fi and fixed-modulation Li-Fi systems. The Li-Fi architecture with AI support possesses a unique edge in intelligent home applications because of its scalability and reliability compared to previous, less dependable wireless networks. This work promises a safe and effective form of data transmission by creating a platform for upcoming advancements in optical wireless communication based on AI. The Li-Fi system proposed that is supported by AI works far better on smart home wireless networking problems of low speeds, insufficient customization, and security. The optimized transmission parameters were used in real-time by the system through a DRL adaptive modulation approach, and this resulted in reducing the bit error rate to 0.00034 and boosting the throughput to 986.45 Mbps.
FPGA-based High-precision MCU testing using pattern latching and DDLL
(IEEE, 2025-07-25) Arunkumar, K; Vani, R; Ali, Guma
Performance testing of microcontrollers (MCUs) is very crucial especially when it comes to safety critical systems such as those in automotive electronics field that will require high accuracy and reliability. One of the essential criteria for performance testing are results checking for getting the maximum possible clock frequency of the MCU possible to influence the automobile units’ efficiency. The traditional method of testing using the ring oscillator approach might not be sufficient in giving MCU performance. To this, the system under consideration includes a pattern latching algorithm with an included configurable ring oscillator. The utilization of a clocked latch with a synchronous gate is used to synchronize ring oscillator circuits so as to do precise pattern analysis. The system makes use of a digital delay locked loop (DDLL) to provide high accuracy when measuring performance by dynamically compensating for delays. Such a strategy increases the validity of performance screening, with improved estimation of the MCU’s operating abilities in adverse conditions.
Scalable multi-connectivity approaches for AR/VR traffic management in 5G networks
(IEEE, 2025-07-15) Dhamini, V.; Subathra, Y.; Aruna, V.; Jothi, J Salomi Backia; Sathish, A.; Ali, Guma
The expanding use of AR/VR applications in 5G networks needs an efficient traffic management system for reaching the network requirements concerning latency and bandwidth usage. The current network architectures show limitations during changes in traffic flow patterns which results in performance deterioration. A Scalable Multi-Connectivity Approach based on Software-Defined Networking (SDN) with Artificial Intelligence (AI) traffic balancing and Multi-Access Edge Computing (MEC) serves to improve 5G network management of AR/VR traffic. The system implementation includes the combination of AI-based load balancing and QoS-aware network slicing and SDN-based adaptive routing for distributing traffic optimally across 5G, LTE, and Wi-Fi networks. Real-time immersive experiences improve significantly through simulation results which show a reduction in AR/VR jitter by 28 % together with an average latency of 15.3 ms. The study shows that artificial intelligence-based multi-connectivity methods create successful traffic management for AR/VR applications through optimized resource utilization in networks.
IoT-enabled smart farming: Harnessing LoRa technology for sustainable agriculture
(IEEE, 2025-07-15) Kalaivanan, T.; Sethuraman, Priya; Abirami, B.; Rajkumar, L.; Sathish, A.; Ali, Guma
The present-day speed of Internet of Things (IoT) technology development delivers real-time monitoring and decision-making features to precision agriculture. This research seeks to establish an IoT-smart farming system built around LoRa (Long Range) technology to create more efficient resource management while maintaining sustainable farming routines. The LoRaWAN-based sensor network distributed across different topographical areas measured environmental parameters such as soil moisture and temperature, humidity. The obtained sensor data was sent through LoRa gateways to establish real-time processing at a cloud server. The radio signals maintained peak power levels in open agricultural fields yet they faced minor reductions in receiving strength in dense farmland regions. Experimental data revealed that LoRa established reliable communication which delivered -70 dBm average RSSI in open fields coupled with -85 dBm RSSI in dense crop areas as well as achievement of over 90% data transmission success across a 5 km distance using minimal power. Results from experiments proved that LoRa technology establishes dependable long-distance transmission while using minimal energy.

Browse

Recent Submissions