View source for Chen2019deepinspect

{{publication
|author=Huili Chen and Cheng Fu and Jishen Zhao and Farinaz Koushanfar
|abstract=<p><span class="fontstyle0">Deep Neural Networks (DNNs) are vulnerable to\&nbsp;</span><span class="fontstyle2">Neural Trojan (NT) </span><span class="fontstyle0">attacks where the adversary injects malicious behaviors during DNN training. This type of \&lsquo;backdoor\&rsquo; attack is activated when the input is stamped with the </span><span class="fontstyle2">trigger </span><span class="fontstyle0">pattern specified by the attacker, resulting in an incorrect prediction of the model. Due to the wide application of DNNs in various critical fields, it is indispensable to inspect whether the pre-trained DNN has been trojaned before employing a model. Our goal in this paper is to address the security concern on unknown DNN to NT attacks and ensure safe model deployment. We propose DeepInspect, the first\&nbsp;</span><span class="fontstyle2">black-box </span><span class="fontstyle0">Trojan detection solution with minimal prior knowledge of the model. DeepInspect learns the probability distribution of potential triggers from the queried model using a </span><span class="fontstyle2">conditional generative model</span><span class="fontstyle0">, thus retrieves the footprint of backdoor insertion. In addition to NT detection, we show that DeepInspect\&rsquo;s trigger generator enables effective Trojan mitigation by model patching. We corroborate the effectiveness, efficiency, and scalability of DeepInspect against the state-of-the-art NT attacks across various benchmarks. Extensive experiments show that DeepInspect offers superior detection performance and lower runtime overhead than the prior work.\&nbsp;</span><br style="font-variant-numeric: normal; font-variant-east-asian: normal; line-height: normal; text-align: -webkit-auto; text-size-adjust: auto;" />
\&nbsp;</p>
|address=Macao, China
|month=8
|year=2019
|booktitle=International Joint Conference on Artificial Intelligence (IJCAI)
|title=DeepInspect: A Black-box Trojan Detection and Mitigation Framework for Deep Neural Networks
|entry=inproceedings
|pdf=Chen2019deepinspect.pdf
}}