Detecting and Eliminating Neural Network Backdoors Through Active Paths with Application to Intrusion Detection
This addresses security vulnerabilities in machine learning models for intrusion detection, though it appears incremental as it builds on existing backdoor detection challenges.
The paper tackles the problem of detecting and eliminating backdoor triggers in neural networks, presenting a novel explainable approach based on active paths and showing promising experimental results in intrusion detection applications.
Machine learning backdoors have the property that the machine learning model should work as expected on normal inputs, but when the input contains a specific $\textit{trigger}$, it behaves as the attacker desires. Detecting such triggers has been proven to be extremely difficult. In this paper, we present a novel and explainable approach to detect and eliminate such backdoor triggers based on active paths found in neural networks. We present promising experimental evidence of our approach, which involves injecting backdoors into a machine learning model used for intrusion detection.