ACL Findings2024
Fooling the Textual Fooler via Randomizing Latent Representations
Duy C. Hoang, Nguyen Hung-Quang, Saurav Manchanda, Minlong Peng, Kok-Seng Wong, Khoa D. Doan
Abstract
Despite outstanding performance in a variety of Natural Language Processing (NLP) tasks, recent studies have revealed that NLP models are vulnerable to adversarial attacks that slightly perturb the input to cause the models to misbehave. Several attacks can even compromise the model without requiring access to the model architecture or model parameters (i.e., a black-box setting), and thus are detrimental to existing NLP applications. To perform these attacks, the adversary queries the victim model many times to determine the most important parts in an input text and transform. In this work, we propose a lightweight and attack-agnostic defense whose main goal is to perplex the process of generating an adversarial example in these query-based black-box attacks; that is to fool the textual fooler. This defense, named AdvFooler, works by randomizing the latent representation of the input at inference time.
Different from existing defenses, AdvFooler does not necessitate additional computational overhead during training nor does it rely on assumptions about the potential adversarial perturbation set while having a negligible impact on the model's accuracy. Our theoretical and empirical analyses highlight the significance of robustness resulting from confusing the adversary via randomizing the latent space, as well as the impact of randomization on clean accuracy. Finally, we empirically demonstrate near state-of-the-art robustness of AdvFooler against representative adversarial attacks on two benchmark datasets.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- BERT-ATTACK: Adversarial Attack Against BERT Using BERT cites TextBugger: Generating Adversarial Text Against Real-world Applications
- Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment cites TextBugger: Generating Adversarial Text Against Real-world Applications
Builds on19
- Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and EntailmentDi Jin, Zhijing Jin, Joey Tianyi Zhou, Peter SzolovitsAAAI 2020 · 1,333 citations
- Certified Robustness to Adversarial Examples with Differential PrivacyMathias Lécuyer, Vaggelis Atlidakis, Roxana Geambasu, Daniel Hsu et al.S&P 2019 · 1,022 citations
- TextBugger: Generating Adversarial Text Against Real-world ApplicationsJinfeng Li, Shouling Ji, Tianyu Du, Bo Li et al.NDSS 2019 · 876 citations
- BERT-ATTACK: Adversarial Attack Against BERT Using BERTLinyang Li, Ruotian Ma, Qipeng Guo, Xiangyang Xue et al.EMNLP 2020 · 529 citations
- FreeLB: Enhanced Adversarial Training for Natural Language UnderstandingChen Zhu, Yu Cheng, Zhe Gan, Siqi Sun et al.ICLR 2020 · 502 citations
Related papers
- "That Is a Suspicious Reaction!": Interpreting Logits Variation to Detect NLP Adversarial AttacksEdoardo Mosca, Shreyash Agarwal, Javier Rando-Ramirez, Georg GrohACL 2022 · 43 citations
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial AttacksNguyen Hung-Quang, Yingjie Lao, Tung Pham, Kok-Seng Wong et al.ICLR 2024 · 3 citations
- Text Adversarial Purification as Defense against Adversarial AttacksLinyang Li, Demin Song, Xipeng QiuACL 2023 · 11 citations
- Disentangled Information Bottleneck for Adversarial Text DefenseYidan Xu, Xinghao Yang, Wei Liu, Bao-di Liu et al.EMNLP 2025
- Weight Perturbation as Defense against Adversarial Word SubstitutionsJianhan Xu, Linyang Li, Jiping Zhang, Xiaoqing Zheng et al.EMNLP Findings 2022 · 5 citations
