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

Download:

Current browse context:

cs.CR

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(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: Diagnosing Vulnerability of Variational Auto-Encoders to Adversarial Attacks

Abstract: In this work, we explore adversarial attacks on the Variational Autoencoders (VAE). We show how to modify data point to obtain a prescribed latent code (supervised attack) or just get a drastically different code (unsupervised attack). We examine the influence of model modifications ($\beta$-VAE, NVAE) on the robustness of VAEs and suggest metrics to quantify it.
Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:2103.06701 [cs.CR]
  (or arXiv:2103.06701v3 [cs.CR] for this version)

Submission history

From: Anna Kuzina [view email]
[v1] Wed, 10 Mar 2021 14:23:20 GMT (5951kb,D)
[v2] Fri, 19 Mar 2021 07:02:22 GMT (5951kb,D)
[v3] Thu, 6 May 2021 08:41:15 GMT (5951kb,D)

Link back to: arXiv, form interface, contact.