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/

Zendy Plus

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

Zendy Tools

Zendy Open

Recent research efforts on adversarial machine learning (ML) haveinvestigated problem-space attacks, focusing on the generation of real evasiveobjects in domains where, unlike images, there is no clear inverse mapping tothe feature space (e.g., software). However, the design, comparison, andreal-world implications of problem-space attacks remain underexplored. Thisarticle makes three major contributions. Firstly, we propose a generalformalization for adversarial ML evasion attacks in the problem-space, whichincludes the definition of a comprehensive set of constraints on availabletransformations, preserved semantics, absent artifacts, and plausibility. Weshed light on the relationship between feature space and problem space, and weintroduce the concept of side-effect features as the by-product of the inversefeature-mapping problem. This enables us to define and prove necessary andsufficient conditions for the existence of problem-space attacks. Secondly,building on our general formalization, we propose a novel problem-space attackon Android malware that overcomes past limitations in terms of semantics andartifacts. We have tested our approach on a dataset with 150K Android apps from2016 and 2018 which show the practical feasibility of evading astate-of-the-art malware classifier along with its hardened version. Thirdly,we explore the effectiveness of adversarial training as a possible approach toenforce robustness against adversarial samples, evaluating its effectiveness onthe considered machine learning models under different scenarios. Our resultsdemonstrate that "adversarial-malware as a service" is a realistic threat, aswe automatically generate thousands of realistic and inconspicuous adversarialapplications at scale, where on average it takes only a few minutes to generatean adversarial instance.

Intriguing Properties of Adversarial ML Attacks in the Problem Space [Extended Version]