Backdoor Attacks and Countermeasures on Deep Learning: A Comprehensive Review

TL;DR

This comprehensive review analyzes various backdoor attack methods on deep learning models, categorizes existing countermeasures, and highlights the need for more effective defenses and future research directions.

Contribution

It systematically categorizes backdoor attacks and defenses, providing a detailed comparison and identifying gaps in current security measures for deep learning.

Findings

Attack surfaces are wide and categorized into six types.

Current defenses are insufficient against adaptive attacks.

Future research should focus on empirical evaluations and practical countermeasures.

Abstract

This work provides the community with a timely comprehensive review of backdoor attacks and countermeasures on deep learning. According to the attacker's capability and affected stage of the machine learning pipeline, the attack surfaces are recognized to be wide and then formalized into six categorizations: code poisoning, outsourcing, pretrained, data collection, collaborative learning and post-deployment. Accordingly, attacks under each categorization are combed. The countermeasures are categorized into four general classes: blind backdoor removal, offline backdoor inspection, online backdoor inspection, and post backdoor removal. Accordingly, we review countermeasures, and compare and analyze their advantages and disadvantages. We have also reviewed the flip side of backdoor attacks, which are explored for i) protecting intellectual property of deep learning models, ii) acting as a honeypot to catch adversarial example attacks, and iii) verifying data deletion requested by the data contributor.Overall, the research on defense is far behind the attack, and there is no single defense that can prevent all types of backdoor attacks. In some cases, an attacker can intelligently bypass existing defenses with an adaptive attack. Drawing the insights from the systematic review, we also present key areas for future research on the backdoor, such as empirical security evaluations from physical trigger attacks, and in particular, more efficient and practical countermeasures are solicited.