Hengzhi Pei (UIUC), Jinyuan Jia (UIUC, Penn State), Wenbo Guo (UC Berkeley, Purdue University), Bo Li (UIUC), Dawn Song (UC Berkeley)

Backdoor attacks have become a major security threat for deploying machine learning models in security-critical applications. Existing research endeavors have proposed many defenses against backdoor attacks. Despite demonstrating certain empirical defense efficacy, none of these techniques could provide a formal and provable security guarantee against arbitrary attacks. As a result, they can be easily broken by strong adaptive attacks, as shown in our evaluation. In this work, we propose TextGuard, the first provable defense against backdoor attacks on text classification. In particular, TextGuard first divides the (backdoored) training data into sub-training sets, achieved by splitting each training sentence into sub-sentences. This partitioning ensures that a majority of the sub-training sets do not contain the backdoor trigger. Subsequently, a base classifier is trained from each sub-training set, and their ensemble provides the final prediction. We theoretically prove that when the length of the backdoor trigger falls within a certain threshold, TextGuard guarantees that its prediction will remain unaffected by the presence of the triggers in training and testing inputs. In our evaluation, we demonstrate the effectiveness of TextGuard on three benchmark text classification tasks, surpassing the certification accuracy of existing certified defenses against backdoor attacks. Furthermore, we propose additional strategies to enhance the empirical performance of TextGuard. Comparisons with state-of-the-art empirical defenses validate the superiority of TextGuard in countering multiple backdoor attacks. Our code and data are available at https://github.com/AI-secure/TextGuard.

View More Papers

WIP: Modeling and Detecting Falsified Vehicle Trajectories Under Data...

Jun Ying, Yiheng Feng (Purdue University), Qi Alfred Chen (University of California, Irvine), Z. Morley Mao (University of Michigan and Google)

Read More

A Comparison of Three Approaches to Assist Users in...

Michael Clark (Brigham Young University), Scott Ruoti (The University of Tennessee), Michael Mendoza (Imperial College London), Kent Seamons (Brigham Young University)

Read More

Understanding and Analyzing Appraisal Systems in the Underground Marketplaces

Zhengyi Li (Indiana University Bloomington), Xiaojing Liao (Indiana University Bloomington)

Read More

Securing EV charging system against Physical-layer Signal Injection Attack...

Soyeon Son (Korea University) Kyungho Joo (Korea University) Wonsuk Choi (Korea University) Dong Hoon Lee (Korea University)

Read More