English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Adversarial robustness of deep models is pivotal in ensuring safe deploymentin real world settings, but most modern defenses have narrow scope andexpensive costs. In this paper, we propose a self-supervised method to detectadversarial attacks and classify them to their respective threat models, basedon a linear model operating on the embeddings from a pre-trainedself-supervised encoder. We use a SimCLR encoder in our experiments, since weshow the SimCLR embedding distance is a good proxy for human perceptibility,enabling it to encapsulate many threat models at once. We call our methodSimCat since it uses SimCLR encoder to catch and categorize various types ofadversarial attacks, including L_p and non-L_p evasion attacks, as well as datapoisonings. The simple nature of a linear classifier makes our method efficientin both time and sample complexity. For example, on SVHN, using only five pairsof clean and adversarial examples computed with a PGD-L_inf attack, SimCat'sdetection accuracy is over 85%. Moreover, on ImageNet, using only 25 examplesfrom each threat model, SimCat can classify eight different attack types suchas PGD-L_2, PGD-L_inf, CW-L_2, PPGD, LPA, StAdv, ReColor, and JPEG-L_inf, withover 40% accuracy. On STL10 data, we apply SimCat as a defense againstpoisoning attacks, such as BP, CP, FC, CLBD, HTBD, halving the success ratewhile using only twenty total poisons for training. We find that the detectorsgeneralize well to unseen threat models. Lastly, we investigate the performanceof our detection method under adaptive attacks and further boost its robustnessagainst such attacks via adversarial training.

Sample Efficient Detection and Classification of Adversarial Attacks via Self-Supervised Embeddings