A Fast Anti-Jamming Algorithm Based on Imitation Learning for WSN.

Area of Science:

• Computer Science
• Electrical Engineering
• Network Security

Background:

• Wireless Sensor Networks (WSNs) are crucial for the Internet of Things (IoT).
• WSNs face increasing threats from malicious jamming attacks.
• Limited hardware resources in WSNs hinder the implementation of complex anti-jamming solutions like Deep Reinforcement Learning (DRL).

Purpose of the Study:

• To propose a rapid anti-jamming method for resource-constrained WSNs.
• To address the challenge of implementing intelligent anti-jamming algorithms in low-cost WSNs.
• To enable efficient anti-jamming strategy acquisition for newly joining network nodes.

Main Methods:

• Developed an imitation learning-based anti-jamming approach.
• Expert anti-jamming trajectories were generated using heuristic algorithms incorporating historical data.
• A Recurrent Neural Network (RNN) was trained to mimic expert decision-making policies.
• Anti-jamming network parameters were transferred to late-access nodes to avoid redundant learning.

Main Results:

• The imitation learning algorithm enabled late-access nodes to rapidly acquire effective anti-jamming strategies.
• Performance of the learned strategies matched expert-level performance.
• The proposed method outperformed traditional Q-learning and Random Frequency Hopping (RFH) algorithms.

Conclusions:

• Imitation learning offers an efficient solution for anti-jamming in resource-limited WSNs.
• The proposed method allows new nodes to quickly achieve expert-level anti-jamming capabilities.
• This approach enhances the resilience of WSNs against jamming attacks without requiring extensive onboard computation.