Abstract

Real-time data offloading is increasingly required in dynamic Internet of Things (IoT) and mobile network environments to reduce latency and enhance computational efficiency. However, this process exposes sensitive information to significant threats, including data leakage, unauthorized access, and adversarial manipulation. To address these challenges, this study proposes a novel Federated Game Theory-Optimized Multi-Layer Perceptron–Long Short-Term Memory (FGO-MLP-LSTM) framework that offers a secure, privacy-preserving, and strategically adaptive solution for real-time data offloading. The core novelty of this work lies in the integration of federated learning, game-theoretic decision modeling, and hybrid deep learning (MLP–LSTM) into a unified security-enhancing architecture. Unlike traditional centralized approaches, the proposed method retains all sensitive data on local edge devices and exchanges only encrypted model updates, substantially mitigating the risk of exposure. Game theory is employed to simulate and analyze interactions among benign agents and adversaries, enabling dynamic optimization of defence strategies under varying threat intensities. Within the framework, the MLP module extracts complex high-dimensional patterns, while the LSTM component captures long-term temporal dependencies essential for detecting real-time anomalies. Preprocessing and min–max normalization further improve data consistency, supporting stable and effective training. Experimental results validate the superiority of the proposed system, achieving 99.3% accuracy, 99.5% precision, 99.4% recall, and 99.4% F1-score—consistently outperforming existing models including Bi-LSTM, LSTM, and Fed-Game. The framework also demonstrates enhanced resilience against poisoning and evasion attacks due to the combined strengths of federated learning and strategic optimization. Overall, the FGO-MLP-LSTM framework provides a scalable and robust solution for secure real-time data offloading, offering clear benefits for next-generation IoT and mobile computing ecosystems.