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

Download:

Current browse context:

cs.CR
< prev | next >
new | recent | 2211

Change to browse by:

cs
cs.LG
cs.SD
eess
eess.AS

References & Citations

DBLP - CS Bibliography

listing | bibtex

Bookmark

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

Computer Science > Cryptography and Security

Title: Going In Style: Audio Backdoors Through Stylistic Transformations

Authors: Stefanos Koffas, Luca Pajola, Stjepan Picek, Mauro Conti
(Submitted on 6 Nov 2022 (v1), last revised 2 May 2023 (this version, v3))
Abstract: This work explores stylistic triggers for backdoor attacks in the audio domain: dynamic transformations of malicious samples through guitar effects. We first formalize stylistic triggers - currently missing in the literature. Second, we explore how to develop stylistic triggers in the audio domain by proposing JingleBack. Our experiments confirm the effectiveness of the attack, achieving a 96% attack success rate. Our code is available in this https URL
Comments: Accepted to ICASSP '23 and the first two authors contributed equally
Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Sound (cs.SD); Audio and Speech Processing (eess.AS)
Cite as: arXiv:2211.03117 [cs.CR]
  (or arXiv:2211.03117v3 [cs.CR] for this version)

Submission history

From: Stefanos Koffas [view email]
[v1] Sun, 6 Nov 2022 13:39:45 GMT (50kb,D)
[v2] Fri, 11 Nov 2022 13:19:31 GMT (51kb,D)
[v3] Tue, 2 May 2023 13:12:54 GMT (53kb,D)
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Link back to: arXiv, form interface, contact.