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

As advances in Deep Neural Networks (DNNs) demonstrate unprecedented levelsof performance in many critical applications, their vulnerability to attacks isstill an open question. We consider evasion attacks at testing time againstDeep Learning in constrained environments, in which dependencies betweenfeatures need to be satisfied. These situations may arise naturally in tabulardata or may be the result of feature engineering in specific applicationdomains, such as threat detection in cyber security. We propose a generaliterative gradient-based framework called FENCE for crafting evasion attacksthat take into consideration the specifics of constrained domains andapplication requirements. We apply it against Feed-Forward Neural Networkstrained for two cyber security applications: network traffic botnetclassification and malicious domain classification, to generate feasibleadversarial examples. We extensively evaluate the success rate and performanceof our attacks, compare their improvement over several baselines, and analyzefactors that impact the attack success rate, including the optimizationobjective and the data imbalance. We show that with minimal effort (e.g.,generating 12 additional network connections), an attacker can change themodel's prediction from the Malicious class to Benign and evade the classifier.We show that models trained on datasets with higher imbalance are morevulnerable to our FENCE attacks. Finally, we demonstrate the potential ofperforming adversarial training in constrained domains to increase the modelresilience against these evasion attacks.

FENCE: Feasible Evasion Attacks on Neural Networks in Constrained Environments