This study introduces a blockchain-based lightweight encryption strategy with federated learning to address the scalability and trust concerns of electronic health records (EHR). After implementing lightweight encryption, the EHR data is stored in a decentralized cloud system. The importance of protecting the privacy and security of distant patients’ health records is explored. Now that stakeholders have a secure portal and cloud data is inaccessible, assaults on electronic healthcare records should decrease. The study guarantees full encryption throughout the whole conversation with federated learning. Deprived of the essential for a trusted third party, the system sets up active smart contracts at runtime between the sensor and the data user to facilitate the transfer of EHR data. To ensure that the data is private between the owner and user during the contract’s execution, it employs a very effective proxy re-encryption mechanism with federated learning. To examine the performance of the proposed system, it has been built and deployed on an Ethereum-based testbed. It is observed that the PSNR and MSE of the proposed model are 39 (1.07×) and 229.6 (1.02×) respectively. The entropy of the image is assessed to be 7.8 for the proposed model. This is also compared with existing algorithms and proved to be a secured model.
Urban sensing is a rapidly growing field that involves gathering real-time data from urban environments. These smart sensing systems are pivotal in enhancing urban planning, resource management, and overall quality of life in cities. However, in the context of data collection and network congestion, non-negligible latency can be experienced in urban environments. Recently, federated learning (FL) has emerged as an attractive approach to address network latency concerns. FL achieves distributing data processing tasks in urban sensing by uploading training results rather than raw data to a central server. In light of this, our paper proposes a wireless sensor air-ground integrated federated learning (AGIFL) network with covert communication. The AGIFL system utilizes a high-altitude platform (HAP) as a parameter server to provide mobility and sustainability of urban sensors and a friendly jammer unmanned aerial vehicle (UAV) to enhance information security. Our main objective is to minimize FL latency for the urban sensor network and ensure secure transmission. To achieve this, we jointly optimize the local training accuracy, interference power, and transmit power of FL sensors. To tackle the nonconvexity of the problem, we decompose the original nonconvex problem into subproblems and solve it using successive convex approximation (SCA) with alternating iterations. Simulation results show that the proposed scheme reduces latency in urban sensor networks significantly and ensures the secure transmission of information compared to benchmarks.
5 July 2024 in IEEE Transactions on Consumer Electronics
A prominent use case of consumer electronics-based Internet of Things (IoT) applications, focused on smart cities, is connected devices that enable cities to optimize their operations via access to high volumes of sensitive data. Yet, these devices commonly utilize public channels for data access and sharing, requiring consistent communication protocols and an Intrusion Detection System (IDS) with the aid of AI. However, most of them involve high computation and communication costs. They are not fully reliable, either. Also, AI-based IDS solutions are viewed as black boxes because they cannot justify their decisions. To resolve these issues, we have proposed a framework based on explainable artificial intelligence (XAI) for securing consumer IoT applications in smart cities. At the beginning of the protocol execution, the participants exchange authenticated data through the blockchain-based AKA procedure. Meanwhile, we adopt the Python-based Shapley Additive Explanation (SHAP) framework to explain and interpret the core features guiding decision-making. The working model of this framework depicts its validation with recent benchmark methods.
16 October 2023 in IEEE Transactions on Consumer Electronics
In customer-centric communication for intelligent cyber-physical transportation systems (ICTS), the extensive deployment of customer electronics will lead to massive overlapping interference, consequently constraining the network capacity. To solve the problem, we design an interference avoidance resource allocation (IARA) strategy based on a hypergraph in customer-centric communication for ICTS. The overlapping interference relationship between customers is modeled in the interference model by hypergraph theory, the IARA under complex interferences is generalized to a hypergraph vertex strong coloring problem. To achieve IARA, a deep Q-network (DQN) algorithm based on reinforcement learning (RL) is proposed to avoid the conflicts of hypergraph coloring. Moreover, for maximizing the network capacity, a robust allocation algorithm with channel state information (RAA-CSI) is proposed for customer-centric communication in ICTS. Simulation results show that compared with the comparison algorithms, the proposed algorithm can average an increase of 25% in the resource reuse rate and effectively improve network capacity for customer-centric communication in ICTS.
10 June 2024 in IEEE Transactions on Consumer Electronics
The integration of Fog computing and 5G communication may enhance Cyber Physical Systems (CPSs) for effective time identification of cyber attacks among consumer electronics. An unsupervised Intrusion Detection System (IDS) based on Generative Adversarial Networks (GANs) and Recurrent Neural Networks (RNNs) is presented in this study. This system is tailored to the resource limitations of consumer electronics within CPSs. By leveraging the processing power of fog nodes and the low-latency capabilities of 5G networks, cyber attacks can be swiftly identified. The use of a trained Long Short-Term Memory (LSTM) network encoder improves detection rates by enhancing reconstruction loss computation. Experimental results demonstrate that this approach, implemented through distributed fog computing infrastructure, offers better detection rates with a 15.2% reduction in detection latency and a 24.2% decrease in overall energy consumption compared to baseline methods. This innovative system could serve as an effective alternative for securing consumer electronic devices integrated into CPSs.
10 January 2025 in IEEE Journal of Biomedical and Health Informatics
The progress in Natural Language Processing (NLP) using Large Language Models (LLMs) has greatly improved medical sentiment analysis of patient feedback extraction from health-related question and answer. However, using LLMs to analyze such data often requires significant training data and computational resources, resulting in considerable increases in training costs and durations, which is one of the primary issues in applying LLMs to real-world healthcare scenarios. To tackle these challenges, a novel optimization algorithm named KAdam-EnGPT4LLM, based on Kalman filters and Adaptive Moment Estimation, is proposed to enhance training efficiency and reduce training costs of LLMs for analyzing patient feedback sentiment. Furthermore, the optimization algorithm KAdam-EnGPT4LLM is employed in training the LLM model GPT4ALL for medical sentiment analysis, resulting in the development of GPT4ALL-MediSentAly-KAdam, which leds to faster convergence and more stable training specifically for medical questions and answer in the context of healthcare. The results show that our GPT4ALL-MediSentAly-KAdam with the optimization algorithm KAdam-EnGPT4LLM achieved better performance that include the best Accuracy, Recall, F1-score, and Runtime for both datasets, outperforming traditional fine-tuned LLMs such as the classic GPT4ALL, Ada, Babbage, Curie, and Duvinci.