Defending Against Unforeseen Failure Modes with Latent Adversarial Training
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911080084668416 |
|---|---|
| author | Casper, Stephen Schulze, Lennart Patel, Oam Hadfield-Menell, Dylan |
| author_facet | Casper, Stephen Schulze, Lennart Patel, Oam Hadfield-Menell, Dylan |
| contents | Despite extensive diagnostics and debugging by developers, AI systems sometimes exhibit harmful unintended behaviors. Finding and fixing these is challenging because the attack surface is so large -- it is not tractable to exhaustively search for inputs that may elicit harmful behaviors. Red-teaming and adversarial training (AT) are commonly used to improve robustness, however, they empirically struggle to fix failure modes that differ from the attacks used during training. In this work, we utilize latent adversarial training (LAT) to defend against vulnerabilities without leveraging knowledge of what they are or using inputs that elicit them. LAT makes use of the compressed, abstract, and structured latent representations of concepts that the network actually uses for prediction. Here, we use it to defend against failure modes without examples that elicit them. Specifically, we use LAT to remove backdoors and defend against held-out classes of adversarial attacks. We show in image classification, text classification, and text generation tasks that LAT usually improves both robustness to novel attacks and performance on clean data relative to AT. This suggests that LAT can be a promising tool for defending against failure modes that are not explicitly identified by developers. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_05030 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Defending Against Unforeseen Failure Modes with Latent Adversarial Training Casper, Stephen Schulze, Lennart Patel, Oam Hadfield-Menell, Dylan Cryptography and Security Artificial Intelligence Machine Learning Despite extensive diagnostics and debugging by developers, AI systems sometimes exhibit harmful unintended behaviors. Finding and fixing these is challenging because the attack surface is so large -- it is not tractable to exhaustively search for inputs that may elicit harmful behaviors. Red-teaming and adversarial training (AT) are commonly used to improve robustness, however, they empirically struggle to fix failure modes that differ from the attacks used during training. In this work, we utilize latent adversarial training (LAT) to defend against vulnerabilities without leveraging knowledge of what they are or using inputs that elicit them. LAT makes use of the compressed, abstract, and structured latent representations of concepts that the network actually uses for prediction. Here, we use it to defend against failure modes without examples that elicit them. Specifically, we use LAT to remove backdoors and defend against held-out classes of adversarial attacks. We show in image classification, text classification, and text generation tasks that LAT usually improves both robustness to novel attacks and performance on clean data relative to AT. This suggests that LAT can be a promising tool for defending against failure modes that are not explicitly identified by developers. |
| title | Defending Against Unforeseen Failure Modes with Latent Adversarial Training |
| topic | Cryptography and Security Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2403.05030 |