Walling up Backdoors in Intrusion Detection Systems

TL;DR

This paper explores the challenge of detecting and removing backdoors in deep learning-based intrusion detection systems, proposing visualization techniques and pruning methods to improve security.

Contribution

It introduces visualization approaches for backdoor detection in IDSs and develops pruning-based defenses for decision trees and random forests, demonstrating their effectiveness.

Findings

Visualization aids in identifying backdoors independently of classifiers.

Common defenses fail to remove backdoors in IDSs.

Pruning-based methods effectively remove backdoors in decision trees and random forests.

Abstract

Interest in poisoning attacks and backdoors recently resurfaced for Deep Learning (DL) applications. Several successful defense mechanisms have been recently proposed for Convolutional Neural Networks (CNNs), for example in the context of autonomous driving. We show that visualization approaches can aid in identifying a backdoor independent of the used classifier. Surprisingly, we find that common defense mechanisms fail utterly to remove backdoors in DL for Intrusion Detection Systems (IDSs). Finally, we devise pruning-based approaches to remove backdoors for Decision Trees (DTs) and Random Forests (RFs) and demonstrate their effectiveness for two different network security datasets.