Stealthy Poisoning Attacks Bypass Defenses in Regression Settings
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915826943131648 |
|---|---|
| author | Carnerero-Cano, Javier Muñoz-González, Luis Spencer, Phillippa Lupu, Emil C. |
| author_facet | Carnerero-Cano, Javier Muñoz-González, Luis Spencer, Phillippa Lupu, Emil C. |
| contents | Regression models are widely used in industrial processes, engineering, and in natural and physical sciences, yet their robustness to poisoning has received less attention. When it has, studies often assume unrealistic threat models and are thus less useful in practice. In this paper, we propose a novel optimal stealthy attack formulation that considers different degrees of detectability and show that it bypasses state-of-the-art defenses. We further propose a new methodology based on normalization of objectives to evaluate different trade-offs between effectiveness and detectability. Finally, we develop a novel defense (BayesClean) against stealthy attacks. BayesClean improves on previous defenses when attacks are stealthy and the number of poisoning points is significant. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_22308 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Stealthy Poisoning Attacks Bypass Defenses in Regression Settings Carnerero-Cano, Javier Muñoz-González, Luis Spencer, Phillippa Lupu, Emil C. Machine Learning Artificial Intelligence Cryptography and Security Regression models are widely used in industrial processes, engineering, and in natural and physical sciences, yet their robustness to poisoning has received less attention. When it has, studies often assume unrealistic threat models and are thus less useful in practice. In this paper, we propose a novel optimal stealthy attack formulation that considers different degrees of detectability and show that it bypasses state-of-the-art defenses. We further propose a new methodology based on normalization of objectives to evaluate different trade-offs between effectiveness and detectability. Finally, we develop a novel defense (BayesClean) against stealthy attacks. BayesClean improves on previous defenses when attacks are stealthy and the number of poisoning points is significant. |
| title | Stealthy Poisoning Attacks Bypass Defenses in Regression Settings |
| topic | Machine Learning Artificial Intelligence Cryptography and Security |
| url | https://arxiv.org/abs/2601.22308 |