Xiangxiang Chen (Zhejiang University), Peixin Zhang (Singapore Management University), Jun Sun (Singapore Management University), Wenhai Wang (Zhejiang University), Jingyi Wang (Zhejiang University)

Model quantization is a popular technique for deploying deep learning models on resource-constrained environments. However, it may also introduce previously overlooked security risks. In this work, we present QuRA, a novel backdoor attack that exploits model quantization to embed malicious behaviors. Unlike conventional backdoor attacks relying on training data poisoning or model training manipulation, QuRA solely works using the quantization operations. In particular, QuRA first employs a novel weight selection strategy to identify critical weights that influence the backdoor target (with the goal of perserving the model's overall performance in mind). Then, by optimizing the rounding direction of these weights, we amplify the backdoor effect across model layers without degrading accuracy. Extensive experiments demonstrate that QuRA achieves nearly 100% attack success rates in most cases, with negligible performance degradation. Furthermore, we show that QuRA can adapt to bypass existing backdoor defenses, underscoring its threat potential. Our findings highlight critical vulnerability in widely used model quantization process, emphasizing the need for more robust security measures. Our implementation is available at https://github.com/cxx122/QuRA.

View More Papers

PhantomMotion: Laser-Based Motion Injection Attacks on Wireless Security Surveillance...

Yan He (University of Oklahoma), Guanchong Huang (University of Oklahoma), Song Fang (University of Oklahoma)

Read More

PIRANHAS: PrIvacy-Preserving Remote Attestation in Non-Hierarchical Asynchronous Swarms

Jonas Hofmann (Technische Universität Darmstadt), Philipp-Florens Lehwalder (Technische Universität Darmstadt), Shahriar Ebrahimi (Alan Turing Institute), Parisa Hassanizadeh (IPPT PAN / University of Warwick), Sebastian Faust (Technische Universität Darmstadt)

Read More

Automating Firmware Vulnerability Triage via High-Level Representations and Similarity...

Daniel Huici, Ricardo J. Rodríguez (University of Zaragoza), Andrei Costin (University of Jyvaskyla), Narges Yousefnezhad (Binare Oy)

Read More