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

There is great potential for damage from adversarial learning (AL) attacks onmachine-learning based systems. In this paper, we provide a contemporary surveyof AL, focused particularly on defenses against attacks on statisticalclassifiers. After introducing relevant terminology and the goals and range ofpossible knowledge of both attackers and defenders, we survey recent work ontest-time evasion (TTE), data poisoning (DP), and reverse engineering (RE)attacks and particularly defenses against same. In so doing, we distinguishrobust classification from anomaly detection (AD), unsupervised fromsupervised, and statistical hypothesis-based defenses from ones that do nothave an explicit null (no attack) hypothesis; we identify the hyperparameters aparticular method requires, its computational complexity, as well as theperformance measures on which it was evaluated and the obtained quality. Wethen dig deeper, providing novel insights that challenge conventional AL wisdomand that target unresolved issues, including: 1) robust classification versusAD as a defense strategy; 2) the belief that attack success increases withattack strength, which ignores susceptibility to AD; 3) small perturbations fortest-time evasion attacks: a fallacy or a requirement?; 4) validity of theuniversal assumption that a TTE attacker knows the ground-truth class for theexample to be attacked; 5) black, grey, or white box attacks as the standardfor defense evaluation; 6) susceptibility of query-based RE to an AD defense.We also discuss attacks on the privacy of training data. We then presentbenchmark comparisons of several defenses against TTE, RE, and backdoor DPattacks on images. The paper concludes with a discussion of future work.

Adversarial Learning in Statistical Classification: A Comprehensive Review of Defenses Against Attacks