Wanlun Ma (Swinburne University of Technology), Derui Wang (CSIRO’s Data61), Ruoxi Sun (The University of Adelaide & CSIRO's Data61), Minhui Xue (CSIRO's Data61), Sheng Wen (Swinburne University of Technology), Yang Xiang (Digital Research & Innovation Capability Platform, Swinburne University of Technology)

Deep Neural Networks (DNNs) are susceptible to backdoor attacks during training. The model corrupted in this way functions normally, but when triggered by certain patterns in the input, produces a predefined target label. Existing defenses usually rely on the assumption of the universal backdoor setting in which poisoned samples share the same uniform trigger. However,
recent advanced backdoor attacks show that this assumption is no longer valid in dynamic backdoors where the triggers vary from input to input, thereby defeating the existing defenses.

In this work, we propose a novel technique, Beatrix (backdoor detection via Gram matrix). Beatrix utilizes Gram matrix to capture not only the feature correlations but also the appropriately high-order information of the representations. By learning class-conditional statistics from activation patterns of normal samples, Beatrix can identify poisoned samples by capturing the anomalies in activation patterns. To further improve the performance in identifying target labels, Beatrix leverages kernel-based testing without making any prior assumptions on representation distribution. We demonstrate the effectiveness of our method through extensive evaluation and comparison with state-of-the-art defensive techniques. The experimental results show that our approach achieves an F1 score of 91.1% in detecting dynamic backdoors, while the state of the art can only reach 36.9%.

View More Papers

A Security Study about Electron Applications and a Programming...

Zihao Jin (Microsoft Research and Tsinghua University), Shuo Chen (Microsoft Research), Yang Chen (Microsoft Research), Haixin Duan (Tsinghua University and...

Read More

Measuring Messengers: Analyzing Infrastructures and Message Timings to Extract...

Theodor Schnitzler (Research Center Trustworthy Data Science and Security, TU Dortmund, and Ruhr-Universität Bochum)

Read More

Assessing the Impact of Interface Vulnerabilities in Compartmentalized Software

Hugo Lefeuvre (The University of Manchester), Vlad-Andrei Bădoiu (University Politehnica of Bucharest), Yi Chen (Rice University), Felipe Huici (Unikraft.io), Nathan...

Read More

Preventing SIM Box Fraud Using Device Model Fingerprinting

BeomSeok Oh (KAIST), Junho Ahn (KAIST), Sangwook Bae (KAIST), Mincheol Son (KAIST), Yonghwa Lee (KAIST), Min Suk Kang (KAIST), Yongdae...

Read More