DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Data
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910896240984064 |
|---|---|
| author | Popovic, Dorde Sadeghi, Amin Yu, Ting Chawla, Sanjay Khalil, Issa |
| author_facet | Popovic, Dorde Sadeghi, Amin Yu, Ting Chawla, Sanjay Khalil, Issa |
| contents | Backdoor attacks are among the most effective, practical, and stealthy attacks in deep learning. In this paper, we consider a practical scenario where a developer obtains a deep model from a third party and uses it as part of a safety-critical system. The developer wants to inspect the model for potential backdoors prior to system deployment. We find that most existing detection techniques make assumptions that are not applicable to this scenario. In this paper, we present a novel framework for detecting backdoors under realistic restrictions. We generate candidate triggers by deductively searching over the space of possible triggers. We construct and optimize a smoothed version of Attack Success Rate as our search objective. Starting from a broad class of template attacks and just using the forward pass of a deep model, we reverse engineer the backdoor attack. We conduct extensive evaluation on a wide range of attacks, models, and datasets, with our technique performing almost perfectly across these settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_21305 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Data Popovic, Dorde Sadeghi, Amin Yu, Ting Chawla, Sanjay Khalil, Issa Cryptography and Security Artificial Intelligence Backdoor attacks are among the most effective, practical, and stealthy attacks in deep learning. In this paper, we consider a practical scenario where a developer obtains a deep model from a third party and uses it as part of a safety-critical system. The developer wants to inspect the model for potential backdoors prior to system deployment. We find that most existing detection techniques make assumptions that are not applicable to this scenario. In this paper, we present a novel framework for detecting backdoors under realistic restrictions. We generate candidate triggers by deductively searching over the space of possible triggers. We construct and optimize a smoothed version of Attack Success Rate as our search objective. Starting from a broad class of template attacks and just using the forward pass of a deep model, we reverse engineer the backdoor attack. We conduct extensive evaluation on a wide range of attacks, models, and datasets, with our technique performing almost perfectly across these settings. |
| title | DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Data |
| topic | Cryptography and Security Artificial Intelligence |
| url | https://arxiv.org/abs/2503.21305 |