RL-Based Frequency Hopping With Block-Shifted Patterns: Balancing Between Anti-Jamming Performance and Synchronization Overhead
Jianshu Zhang, Xiaofu Wu
Abstract
Wireless transmissions are inherently vulnerable to jamming attacks. Frequency hopping (FH) has been widely employed in practice to combat jamming attacks. Although the recent use of reinforcement learning (RL) could improve the anti-jamming performance, a dynamic selection of working frequency in RL-based FH schemes makes them less attractive due to significant overhead for achieving the frequency synchronization between legitimate transmitter and receiver. In order to get a balance between anti-jamming performance and synchronization overhead, this article proposes a novel RL-based FH approach with block-shifted patterns, where the FH pattern between legitimate transmitter and receiver slowly changes in a block-wise manner and the pattern under a block window is shifted with an offset determined by an RL algorithm. In this way, the synchronization overhead could be well alleviated. Then, we derive some theoretical results for the effect of block size on the anti-jamming performance. Extensive experiments show that the proposed algorithm can reach the analytic bound on throughput of the proposed FH scheme. By properly selecting a block size, the proposed FH scheme could achieve the target balance between anti-jamming performance and synchronization overhead.