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

Numerous adversarial attack methods have been developed to generateimperceptible image perturbations that can cause erroneous predictions ofstate-of-the-art machine learning (ML) models, in particular, deep neuralnetworks (DNNs). Despite intense research on adversarial attacks, little effortwas made to uncover 'arcana' carried in adversarial attacks. In this work, weask whether it is possible to infer data-agnostic victim model (VM) information(i.e., characteristics of the ML model or DNN used to generate adversarialattacks) from data-specific adversarial instances. We call this 'model parsingof adversarial attacks' - a task to uncover 'arcana' in terms of the concealedVM information in attacks. We approach model parsing via supervised learning,which correctly assigns classes of VM's model attributes (in terms ofarchitecture type, kernel size, activation function, and weight sparsity) to anattack instance generated from this VM. We collect a dataset of adversarialattacks across 7 attack types generated from 135 victim models (configured by 5architecture types, 3 kernel size setups, 3 activation function types, and 3weight sparsity ratios). We show that a simple, supervised model parsingnetwork (MPN) is able to infer VM attributes from unseen adversarial attacks iftheir attack settings are consistent with the training setting (i.e.,in-distribution generalization assessment). We also provide extensiveexperiments to justify the feasibility of VM parsing from adversarial attacks,and the influence of training and evaluation factors in the parsing performance(e.g., generalization challenge raised in out-of-distribution evaluation). Wefurther demonstrate how the proposed MPN can be used to uncover the source VMattributes from transfer attacks, and shed light on a potential connectionbetween model parsing and attack transferability.

Can Adversarial Examples Be Parsed to Reveal Victim Model Information?