How Hard is Trojan Detection in DNNs? Fooling Detectors With Evasive Trojans

Mantas Mazeika, Andy Zou, Akul Arora, Pavel Pleskov, Dawn Song, Dan Hendrycks, Bo Li, David ForsythSubmission to ICLR 2024 · Rejected

Abstract

Trojan attacks can pose serious risks by injecting deep neural networks with hidden, adversarial functionality. Recent methods for detecting whether a model is trojaned appear highly successful. However, a concerning and relatively unexplored possibility is that trojaned networks could be made harder to detect. To better understand the scope of this risk, we develop a general method for making trojans more evasive based on several novel techniques and observations. In experiments, we find that our evasive trojans reduce the efficacy of a wide range of detectors across numerous evaluation settings while maintaining high attack success rates. Surprisingly, we also find that our evasive trojans are substantially harder to reverse-engineer despite not being explicitly designed with this attribute in mind. These findings underscore the importance of developing more robust monitoring mechanisms for hidden functionality and clarifying the offense-defense balance of trojan detection.
BibTeX
@inproceedings{mazeika2023how,
  author = {Mantas Mazeika and Andy Zou and Akul Arora and Pavel Pleskov and Dawn Song and Dan Hendrycks and Bo Li and David Forsyth},
  title = {{How Hard is Trojan Detection in DNNs? Fooling Detectors With Evasive Trojans}},
  booktitle = {International Conference on Learning Representations},
  year = {2023},
  url = {https://openreview.net/forum?id=V-RDBWYf0go},
}

Loading related content...

Figure