AppCon: Mitigating Evasion Attacks to ML Cyber Detectors

Abstract

Adversarial attacks represent a critical issue that prevents the reliable integration of machine learning methods into cyber defense systems. Past work has shown that even proficient detectors are highly affected just by small perturbations to malicious samples, and that existing countermeasures are immature. We address this problem by presenting AppCon, an original approach to harden intrusion detectors against adversarial evasion attacks. Our proposal leverages the integration of ensemble learning to realistic network environments, by combining layers of detectors devoted to monitor the behavior of the applications employed by the organization. Our proposal is validated through extensive experiments performed in heterogeneous network settings simulating botnet detection scenarios, and consider detectors based on distinct machine- and deep-learning algorithms. The results demonstrate the effectiveness of AppCon in mitigating the dangerous threat of adversarial attacks in over 75% of the considered evasion attempts, while not being affected by the limitations of existing countermeasures, such as performance degradation in non-adversarial settings. For these reasons, our proposal represents a valuable contribution to the development of more secure cyber defense platforms.

GADoT: GAN-based Adversarial Training for Robust DDoS Attack Detection

“…Notably, we observe a slight drop in precision on the SYN datasets, and in recall score on the CICIDS2017 dataset. This indicates that a few ambiguous network flows are misclassified, which matches the observations made in [17] after adversarial training and [32] after employing ensemble learning to improve robustness.…”

TAD: Transfer learning-based multi-adversarial detection of evasion attacks against network intrusion detection systems

“…It is worth mentioning that other works have studied the use of ensemble techniques to improve the detection rate against adversarial attacks, such as [17], [14] and [28], although in these papers the use of ensemble techniques has a different meaning than the one we consider in this paper. For them, the ensemble technique is to train each classifier on a specific attack or application to ensure proper tuning for that task, but this might lead to less generalization in the training process, making it difficult to handle unknown attacks.…”

“…The overall scheme for them is to assemble multiple classifiers, each specialized in a particular attack or application, whereas for ours, each instance is inspected by all components of the ensemble scheme leading to a deeper inspection of each particular instance. It should also be noted that approaches [17], [14] and [28] seek to make the classifier robust by adding and/or filtering the training data used to train the model. This is not always possible, especially in the case of an already trained and deployed IDS.…”

Network Attack Detection Method of the Cyber-Physical Power System Based on Ensemble Learning

“…Adversarial attacks will make the system miss or cause a misjudgment. Apruzzese et al [37] address this problem by presenting AppCon, an approach to harden intrusion detectors against adversarial evasion attacks. The proposal leverages the integration of ensemble learning into realistic network environments, improving the detection rate against evasion attacks.…”