Assessing Representation Stability for Transformer Models
By: Bryan E. Tuck, Rakesh M. Verma
Potential Business Impact:
Stops bad words from tricking computers.
Adversarial text attacks remain a persistent threat to transformer models, yet existing defenses are typically attack-specific or require costly model retraining. We introduce Representation Stability (RS), a model-agnostic detection framework that identifies adversarial examples by measuring how embedding representations change when important words are masked. RS first ranks words using importance heuristics, then measures embedding sensitivity to masking top-k critical words, and processes the resulting patterns with a BiLSTM detector. Experiments show that adversarially perturbed words exhibit disproportionately high masking sensitivity compared to naturally important words. Across three datasets, three attack types, and two victim models, RS achieves over 88% detection accuracy and demonstrates competitive performance compared to existing state-of-the-art methods, often at lower computational cost. Using Normalized Discounted Cumulative Gain (NDCG) to measure perturbation identification quality, we reveal that gradient-based ranking outperforms attention and random selection approaches, with identification quality correlating with detection performance for word-level attacks. RS also generalizes well to unseen datasets, attacks, and models without retraining, providing a practical solution for adversarial text detection.
Similar Papers
Towards Trustworthy Wi-Fi Sensing: Systematic Evaluation of Deep Learning Model Robustness to Adversarial Attacks
Machine Learning (CS)
Makes wireless sensing safer from hacking.
RobustMask: Certified Robustness against Adversarial Neural Ranking Attack via Randomized Masking
Cryptography and Security
Protects search results from fake information.
Test-Time Defense Against Adversarial Attacks via Stochastic Resonance of Latent Ensembles
CV and Pattern Recognition
Protects AI from being tricked by fake images.