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

In recent years, a variety of effective neural network-based methods foranomaly and cyber attack detection in industrial control systems (ICSs) havebeen demonstrated in the literature. Given their successful implementation andwidespread use, there is a need to study adversarial attacks on such detectionmethods to better protect the systems that depend upon them. The extensiveresearch performed on adversarial attacks on image and malware classificationhas little relevance to the physical system state prediction domain, which mostof the ICS attack detection systems belong to. Moreover, such detection systemsare typically retrained using new data collected from the monitored system,thus the threat of adversarial data poisoning is significant, however thisthreat has not yet been addressed by the research community. In this paper, wepresent the first study focused on poisoning attacks on online-trainedautoencoder-based attack detectors. We propose two algorithms for generatingpoison samples, an interpolation-based algorithm and a back-gradientoptimization-based algorithm, which we evaluate on both synthetic andreal-world ICS data. We demonstrate that the proposed algorithms can generatepoison samples that cause the target attack to go undetected by the autoencoderdetector, however the ability to poison the detector is limited to a small setof attack types and magnitudes. When the poison-generating algorithms areapplied to the popular SWaT dataset, we show that the autoencoder detectortrained on the physical system state data is resilient to poisoning in the faceof all ten of the relevant attacks in the dataset. This finding suggests thatneural network-based attack detectors used in the cyber-physical domain aremore robust to poisoning than in other problem domains, such as malwaredetection and image processing.

Can't Boil This Frog: Robustness of Online-Trained Autoencoder-Based Anomaly Detectors to Adversarial Poisoning Attacks