The accurate prediction of cryptocurrency prices is crucial due to the volatility and complexity of digital asset markets, which pose significant challenges to traders, investors, and researchers. This research addresses these challenges by leveraging machine learning and deep learning techniques to forecast closing prices for cryptocurrencies, focusing on Bitcoin, Ethereum, Binance Coin, and Litecoin cryptocurrency datasets. A Random Forest ensemble learning algorithm, a Gradient Boosting model, and a feedforward neural network were implemented to handle the complexities in cryptocurrency data. A Z-Score-based anomaly detection framework was integrated to classify closing prices as normal or abnormal, aiding in identifying significant market events. Evaluation metrics, such as the Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²), demonstrate the superior precision and reliability of the Random Forest and Gradient Boosting models. The deep learning model indicates strong generalization capabilities, suggesting potential advantages on more complex datasets. These findings highlight the importance of combining advanced machine learning techniques and cryptocurrencies to develop a robust framework for cryptocurrency forecasting and anomaly detection.

Mobile Edge Computing (MEC) is well known for its critical role in controlling the growth of Internet of Things (IoT) applications, even though energy demand forecasting is inherently challenging because of multiple task-specific factors. Including a microgrid as a backup energy source offers a chance to improve the effectiveness of energy distribution, but it also raises concerns about using non-renewable energy sources. This study explores the complex task of managing energy scheduling in a microgrid-enabled MEC network. It does this by minimizing residual planned energy using the conditional value at risk (CVaR) measure and illuminate the complexity of this NP-hard problem. To address this complex problem, a multi-agent stochastic game framework is implemented to achieve a Nash equilibrium for collective policies and prove model convergence, revealing the novel Asynchronous Federated Learning framework for Microgrid-Enabled MEC Network (AFFIN). A device selection strategy and a multi-agent Asynchronous Advantage Actor-Critic (A3C) algorithm with shared neural networks are employed to enhance learning efficiency. The proposed scheme enables localized model training and parameter sharing with the MEC server. The results demonstrated the exceptional effectiveness of incorporating CVaR into energy scheduling, surpassing the performance of random agent and single-agent models in terms of optimizing energy distribution in the MEC network.

The Internet of Things (IoT) has emerged as an innovative paradigm that interconnects a diverse array of devices and systems to enable uninterrupted communication and data transmission. Within this extensive architecture, device-to-device (D2D) communication is critical for facilitating direct interactions among linked devices. This aspect of IoT surpasses traditional modalities of engagement between humans and machines. It cultivates an interactive ecosystem wherein devices autonomously collaborate, share data, and perform designated functions. D2D communication offers a multitude of benefits, such as enabling short-range interactions, minimizing latency, enhancing scalability, and optimizing energy efficiency. Moreover, to promote widespread connectivity and exceptionally dependable low-latency performance within the fifth generation (5G) network, the adoption of non-orthogonal multiple access (NOMA) merits further exploration. In this regard, the impact of federated learning on NOMA-based D2D group users (DGUs) within wireless-powered communication frameworks is examined. Initially, the D2D transmitters (DDTs) harvest energy from the radio frequency signals emitted by the base station. Subsequently, the DDTs employ NOMA to establish communication with the D2D receivers (DDRs) by utilizing the energy they have accumulated. A stochastic optimization problem is formulated to enhance energy efficiency (EE) and minimize delay. This formulation incorporates both stochastic traffic arrivals and the time-varying conditions of the communication channel. By applying the Markov decision process, the non-convex optimization problem is transformed into a mathematical model that encapsulates decision-making scenarios. Furthermore, federated learning is implemented to achieve the objectives and accelerate the dissemination of local training data across the DGUs. Empirical results illustrate that the proposed methodology achieves performance metrics that are 8.47% and 66.10% superior to those of distributed and centralized schemes, respectively.

Recent studies have shown that sponge attacks can significantly increase the energy consumption and inference latency of deep neural networks (DNNs). However, prior work has focused primarily on computer vision and natural language processing tasks, overlooking the growing use of lightweight AI models in sensing-based applications on resource-constrained devices, such as those in Internet of Things (IoT) environments. These attacks pose serious threats of energy depletion and latency degradation in systems where limited battery capacity and real-time responsiveness are critical for reliable operation. This paper makes two key contributions. First, we present the first systematic exploration of energy-latency sponge attacks targeting sensing-based AI models. Using wearable sensing-based AI as a case study, we demonstrate that sponge attacks can substantially degrade performance by increasing energy consumption, leading to faster battery drain, and by prolonging inference latency. Second, to mitigate such attacks, we investigate model pruning, a widely adopted compression technique for resource-constrained AI, as a potential defense. Our experiments show that pruning-induced sparsity significantly improves model resilience against sponge poisoning. We also quantify the trade-offs between model efficiency and attack resilience, offering insights into the security implications of model compression in sensing-based AI systems deployed in IoT environments.

The development of blockchain technology has facilitated its application in many areas of life due to its ability to enable decentralization in processing and storage. To support blockchain's application in various fields, its algorithms, and consensus protocols should be developed. Nowadays, there are many consensus protocols, each of which can be applied to the appropriate field, considering the cost of each protocol. Users should be motivated to use blockchain and participate in its processes. Considering the processing, communication, and storage costs, effective reward mechanisms are necessary to compensate users and encourage participation. Therefore, this study investigates and reviews a variety of blockchain consensus protocols and corresponding reward strategies. The study analyses twenty-two consensus protocols, explaining how they work, and their advantages and disadvantages. In addition, this paper discusses the types of reward strategies, their sources, limitations, and future. According to the nature of consensus protocols, they are classified into seven categories, which are computational-based, wealth-based, behavioral-based, fairness-based, fault-tolerance-based, leader-based, and transactional-based. The study aims to provide information on the future paths of blockchain consensus mechanisms, promoting a more profound understanding of their capabilities and constraints in many situations.