We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Cryptography and Security

Title: Invisible Backdoor Attack with Sample-Specific Triggers

Abstract: Recently, backdoor attacks pose a new security threat to the training process of deep neural networks (DNNs). Attackers intend to inject hidden backdoors into DNNs, such that the attacked model performs well on benign samples, whereas its prediction will be maliciously changed if hidden backdoors are activated by the attacker-defined trigger. Existing backdoor attacks usually adopt the setting that triggers are sample-agnostic, $i.e.,$ different poisoned samples contain the same trigger, resulting in that the attacks could be easily mitigated by current backdoor defenses. In this work, we explore a novel attack paradigm, where backdoor triggers are sample-specific. In our attack, we only need to modify certain training samples with invisible perturbation, while not need to manipulate other training components ($e.g.$, training loss, and model structure) as required in many existing attacks. Specifically, inspired by the recent advance in DNN-based image steganography, we generate sample-specific invisible additive noises as backdoor triggers by encoding an attacker-specified string into benign images through an encoder-decoder network. The mapping from the string to the target label will be generated when DNNs are trained on the poisoned dataset. Extensive experiments on benchmark datasets verify the effectiveness of our method in attacking models with or without defenses.
Comments: It is accepted by ICCV 2021
Subjects: Cryptography and Security (cs.CR)
Cite as: arXiv:2012.03816 [cs.CR]
  (or arXiv:2012.03816v3 [cs.CR] for this version)

Submission history

From: Yuezun Li [view email]
[v1] Mon, 7 Dec 2020 16:02:08 GMT (8173kb,D)
[v2] Thu, 12 Aug 2021 03:45:40 GMT (9063kb,D)
[v3] Fri, 13 Aug 2021 01:04:15 GMT (9063kb,D)

Link back to: arXiv, form interface, contact.