The obfuscated traffic generation system based on feature payload recombination

• 1、School of Electronic Engineering, Beijing University of Posts and Telecommunications,Beijing 100876
• 2、School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：As data traffic in networks experiences explosive growth and cybersecurity incidents become more frequent, new challenges are posed to network intrusion detection. Artificial intelligence algorithms driven by labeled data have been widely introduced into the field of network intrusion detection, with their robustness becoming a new research focus. We focuses on studying the adversarial robustness of classifiers in dealing with obfuscated traffic. A system is proposed for generating obfuscated traffic based on feature payload recombination, simulating the generation of obfuscated traffic from maliciously labeled traffic. The validation results indicate that obfuscated traffic generated through feature payload recombination exhibits significant effectiveness in actual transmission. Injecting obfuscated traffic into the traffic set can effectively weaken the malicious traffic classification performance of network traffic classifiers.

Keywords： deep learning network traffic obfuscated traffic robustness