Introducing Adaptive Continuous Adversarial Training (ACAT) to Enhance ML Robustness
This addresses the problem of improving ML robustness against adversarial attacks in cybersecurity, specifically for spam detection, but it is incremental as it builds on existing adversarial training methods.
The paper tackled the challenge of costly labeled adversarial training data in cybersecurity by introducing Adaptive Continuous Adversarial Training (ACAT), which integrates adversarial samples during continuous learning, resulting in reduced detection time and increased accuracy from 69% to over 88% for a spam filter after three retraining sessions.
Adversarial training enhances the robustness of Machine Learning (ML) models against adversarial attacks. However, obtaining labeled training and adversarial training data in network/cybersecurity domains is challenging and costly. Therefore, this letter introduces Adaptive Continuous Adversarial Training (ACAT), a method that integrates adversarial training samples into the model during continuous learning sessions using real-world detected adversarial data. Experimental results with a SPAM detection dataset demonstrate that ACAT reduces the time required for adversarial sample detection compared to traditional processes. Moreover, the accuracy of the under-attack ML-based SPAM filter increased from 69% to over 88% after just three retraining sessions.