paper-with-me

Papers

Clean Label Attacks against SLU Systems

2024-09-13 · Henry Li Xinyuan, Sonal Joshi, Thomas Thebaud, Jesus Villalba, Najim Dehak, Sanjeev Khudanpur

Poisoning backdoor attacks involve an adversary manipulating the training data to induce certain behaviors in the victim model by inserting a trigger in the signal at inference time. We adapted clean label backdoor (CLBD)-data poisoning attacks, which do not modify the training labels, on state-of-the-art speech recognition models that support/perform a Spoken Language Understanding task, achieving 99.8% attack success rate by poisoning 10% of the training data. We analyzed how varying the signal-strength of the poison, percent of samples poisoned, and choice of trigger impact the attack. We also found that CLBD attacks are most successful when applied to training samples that are inherently hard for a proxy model. Using this strategy, we achieved an attack success rate of 99.3% by poisoning a meager 1.5% of the training data. Finally, we applied two previously developed defenses against gradient-based attacks, and found that they attain mixed success against poisoning.

📄 PDF Abstract BibTeX arXiv:2409.08985

Code (0)

등록된 구현이 없습니다.

Tasks

Data Poisoningspeech-recognitionSpeech RecognitionSpoken Language Understanding

Similar Papers 제목 키워드 기반

Deep k-NN Defense against Clean-label Data Poisoning Attacks

2019-09-29 · Neehar Peri, Neal Gupta, W. Ronny Huang, Liam Fowl 외

Targeted clean-label data poisoning is a type of adversarial attack on machine learning systems in which an adversary injects a few correctly-labeled, minimally-perturbed samples into the training data, causing a model t…

Adversarial AttackData Poisoning

Certified Robustness to Clean-Label Poisoning Using Diffusion Denoising

2024-03-18 · Sanghyun Hong, Nicholas Carlini, Alexey Kurakin

We present a certified defense to clean-label poisoning attacks under $\ell_2$-norm. These attacks work by injecting a small number of poisoning samples (e.g., 1%) that contain bounded adversarial perturbations into the …

Adversarial RobustnessDenoising

Low-Loss Subspace Compression for Clean Gains against Multi-Agent Backdoor Attacks

2022-03-07 · Siddhartha Datta, Nigel Shadbolt

Recent exploration of the multi-agent backdoor attack demonstrated the backfiring effect, a natural defense against backdoor attacks where backdoored inputs are randomly classified. This yields a side-effect of low accur…

Backdoor Attack

Mirage: a Clean-Label Backdoor against LiDAR 3D Object Detection

2026-06-18 · Ziba Parsons, Ang Li arxiv

Deep neural network-based LiDAR 3D object detection serves as a critical perception component in safety-critical autonomous systems. However, recent studies have revealed its vulnerability to backdoor attacks. Existing a…

3D Object Detection

Checkerboard: Closed-Form and Data-Independent Trigger Design for Clean-Label Backdoor Attacks

2026-05-02 · Yi Yang, Jinyang Huang, Binbin Liu, Feng-Qi Cui 외 arxiv

Backdoor attacks threaten the deep-learning supply chain by poisoning a small fraction of the training data so that a model behaves normally on clean inputs but maps triggered inputs to an attacker-chosen class. Clean-la…