Abstract
Modern integrated sensing and communication (ISAC) technology enables the tight integration of sensing and transmission within the framework of maritime Internet of Things (IoT) to achieve enhanced performance. However, in actual maritime industrial environments, such maritime IoT systems are often susceptible to various sources of interference, particularly anthropogenic jamming, making the application of antijamming ISAC technology in maritime IoT an imperative challenge. To address this issue, this article proposes a novel algorithm integrating dynamic attack detection for solving transmission scheduling optimization problems under Denial-of-Service (DoS) attacks. First, a novel learning-based adaptive Kalman filter (LAKF) algorithm is introduced. Even in time-varying noise environments, the algorithm can still dynamically adjust the covariance matrix of observation noise through reinforcement learning to match the actual noise characteristics. Second, an intelligent detection algorithm is designed to counter time-varying attack patterns, which leverages a temporal window approach incorporating long short-term memory networks and an attention mechanism. Finally, we propose an energy-aware LAKF-attack pattern detection-twin delayed deep deterministic policy gradient tracking (APD-TD3T) algorithm, which integrates the LAKF and APD algorithms and uses the improved TD3 algorithm (an improvement of the DDPG algorithm) as the core algorithm. Specially, it also incorporates a delayed reward design, which effectively addresses the anti-DoS transmission scheduling optimization problem in continuous action space. Simulation experiment and comparison results demonstrate that the proposed LAKF-APD-TD3T algorithm achieves superior performance, significantly decreasing both tracking error and energy consumption in target tracking scenarios.
Funding Information
  • National Natural Science Foundation of China (62303298)