Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees

Anomaly-based attack detection methods depend on some form of machine learning to detect data falsification attacks in smart living cyber–physical systems. However, there is a lack of studies that consider the presence of attacks during the training phase and their effect on detection and false alar...

Full description

Saved in:
Bibliographic Details
Main Authors: Sahar Abedzadeh, Shameek Bhattacharjee
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Information
Subjects:
Online Access:https://www.mdpi.com/2078-2489/16/6/428
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849431806111121408
author Sahar Abedzadeh
Shameek Bhattacharjee
author_facet Sahar Abedzadeh
Shameek Bhattacharjee
author_sort Sahar Abedzadeh
collection DOAJ
description Anomaly-based attack detection methods depend on some form of machine learning to detect data falsification attacks in smart living cyber–physical systems. However, there is a lack of studies that consider the presence of attacks during the training phase and their effect on detection and false alarm performance. To improve the robustness of time series learning for anomaly detection, we propose a framework by modifying design choices such as regression error type and loss function type while learning the thresholds for an anomaly detection framework during the training phase. Specifically, we offer theoretical proofs on the relationship between poisoning attack strengths and how that informs the choice of loss functions used to learn the detection thresholds. This, in turn, leads to explainability of why and when our framework mitigates data poisoning and the trade-offs associated with such design changes. The theoretical results are backed by experimental results that prove attack mitigation performance with NIST-specified metrics for CPS, using real data collected from a smart metering infrastructure as a proof of concept. Thus, the contribution is a framework that guarantees security of ML and ML for security simultaneously.
format Article
id doaj-art-e46faf269d1b49ce8cca90c7509b41d4
institution Kabale University
issn 2078-2489
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Information
spelling doaj-art-e46faf269d1b49ce8cca90c7509b41d42025-08-20T03:27:32ZengMDPI AGInformation2078-24892025-05-0116642810.3390/info16060428Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable GuaranteesSahar Abedzadeh0Shameek Bhattacharjee1Computer Science Department, Western Michigan University, Kalamazoo, MI 49008, USAComputer Science Department, Western Michigan University, Kalamazoo, MI 49008, USAAnomaly-based attack detection methods depend on some form of machine learning to detect data falsification attacks in smart living cyber–physical systems. However, there is a lack of studies that consider the presence of attacks during the training phase and their effect on detection and false alarm performance. To improve the robustness of time series learning for anomaly detection, we propose a framework by modifying design choices such as regression error type and loss function type while learning the thresholds for an anomaly detection framework during the training phase. Specifically, we offer theoretical proofs on the relationship between poisoning attack strengths and how that informs the choice of loss functions used to learn the detection thresholds. This, in turn, leads to explainability of why and when our framework mitigates data poisoning and the trade-offs associated with such design changes. The theoretical results are backed by experimental results that prove attack mitigation performance with NIST-specified metrics for CPS, using real data collected from a smart metering infrastructure as a proof of concept. Thus, the contribution is a framework that guarantees security of ML and ML for security simultaneously.https://www.mdpi.com/2078-2489/16/6/428anomaly detectiondata poisoning attackscyber–physical systems (CPS)machine learning robustnessML for securityresilient learning-based CPS
spellingShingle Sahar Abedzadeh
Shameek Bhattacharjee
Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
Information
anomaly detection
data poisoning attacks
cyber–physical systems (CPS)
machine learning robustness
ML for security
resilient learning-based CPS
title Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
title_full Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
title_fullStr Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
title_full_unstemmed Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
title_short Mitigating Impact of Data Poisoning Attacks on CPS Anomaly Detection with Provable Guarantees
title_sort mitigating impact of data poisoning attacks on cps anomaly detection with provable guarantees
topic anomaly detection
data poisoning attacks
cyber–physical systems (CPS)
machine learning robustness
ML for security
resilient learning-based CPS
url https://www.mdpi.com/2078-2489/16/6/428
work_keys_str_mv AT saharabedzadeh mitigatingimpactofdatapoisoningattacksoncpsanomalydetectionwithprovableguarantees
AT shameekbhattacharjee mitigatingimpactofdatapoisoningattacksoncpsanomalydetectionwithprovableguarantees