Design and Detection of Covert Man-in-the-Middle Cyberattacks on Water Treatment Plants Conference: Proceedings of the 2025 Workshop on Re-design Industrial Control Systems with Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 Workshop on Re-design Industrial Control Systems wit
Victor Mattos, João Henrique Schmidt, Amit Bhaya, Alan Oliveira de Sá, Daniel Sadoc Menasché, Gaurav Srivastava · 2025
Read Design and Detection of Covert Man-in-the-Middle Cyberattacks on Water Treatment Plants Conference: Proceedings of the 2025 Workshop on Re-design Industrial Control Systems with Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 Workshop on Re-design Industrial Control Systems wit by Victor Mattos, João Henrique Schmidt, Amit Bhaya, Alan Oliveira de Sá, Daniel Sadoc Menasché, Gaurav Srivastava (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-03971.txt.
Introduction
Victor Mattos, Joao Henrique Schmidt, Amit Bhaya, Alan Oliveira de Sa, Daniel Sadoc Menasche, and Gaurav Srivastava model covert man-in-the-middle attacks on water treatment plants. Digital intrusion there can threaten health, safety, and the environment. Their systematic approach uses system identification to design an attacker's covert controller.
A second-order linear model with time delay stands in for plant dynamics, and stealthy attacks are simulated against a process-aware anomaly detector. System noise and errors in the attacker's own plant model change how stealthy an attack can stay. Those sensitivities, the authors argue, show why industrial control defences need stronger detection built for quiet, model-aware intruders.
Catalog record
| CREATOR | Victor Mattos, João Henrique Schmidt, Amit Bhaya, Alan Oliveira de Sá, Daniel Sadoc Menasché, Gaurav Srivastava |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-03971
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Design and Detection of Covert Man-in-the-Middle Cyberattacks on Water Treatment Plants Conference: Proceedings of the 2025 Workshop on Re-design Industrial Control Systems with Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 Workshop on Re-design Industrial Control Systems with Security (RICSS ’25), October 13–17, 2025, Taipei, Taiwan DOI: 10.1145/3733823.3764516 ISBN: 979-8-4007-1911-0/2025/10 Thanks: This work was partially supported by CAPES, Finance Code 001; by FAPERJ under grant E-26/204.268/2024; and by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil, under grants 403601/2023-1, 315106/2023-9, and BPP/PQ-Sr 313335/2022-2, as well as FCT through the LASIGE Research Unit, ref. UID/00408/2025. Victor Mattos, João Henrique Schmidt, Amit Bhaya, Alan Oliveira de Sá, Daniel Sadoc Menasché, Gaurav Srivastava Abstract . Cyberattacks targeting critical infrastructure — such as water treatment facilities — represent significant threats to public health, safety, and the environment. This paper introduces a systematic approach for modeling and assessing covert man-in-the-middle (MitM) attacks that leverage system identification techniques to inform the attack design. We focus on the attacker’s ability to deploy a covert controller, and we evaluate countermeasures based on the Process-Aware Stealthy Attack Detection (PASAD) anomaly detection method. Using a second-order linear time-invariant with time delay model, representative of water treatment dynamics, we design and simulate stealthy attacks. Our results highlight how factors such as system noise and inaccuracies in the attacker’s plant model influence the attack’s stealthiness, underscoring the need for more robust detection strategies in industrial control environments. 1. Introduction Cyber-physical systems (CPS), such as those found in industrial control and water treatment plants, are increasingly exposed to sophisticated adversaries capable of launching stealthy cyberattacks. Among these, covert man-in-the-middle (MitM) attacks pose a particular challenge: by manipulating both control commands and sensor measurements, attackers can alter system behavior without triggering alarms. The effectiveness of such attacks is amplified when the adversary first engages in system identification (SI) to estimate the plant and controller dynamics, and then uses that information to minimize the perception of the effect caused by the attack ( 9 ) . 1 1 1 This is an extended version of ( 17 ) . © ACM 2025. This is the author’s version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in the Proceedings of the 2025 Workshop on Re-design Industrial Control Systems with Security (ACM CCS Workshop), https://doi.org/10.1145/3733823.3764516. This paper investigates the design of covert MitM attacks that leverage SI techniques to learn the system’s input-output behavior. Once the attacker has an approximate model, they can predict the system’s response to malicious inputs and inject compensatory signals to cancel their observable effects, effectively concealing the attack ( 26 ) . We assume that the SI phase has already been completed and focus on the second stage: designing attack strategies and using detection mechanisms to evaluate the success of the attack. In particular, we explore the limits of stealth under different model accuracies and detection thresholds and also assess the performance of advanced anomaly detectors in identifying these hidden manipulations. Note that considering the attack strategy as well as the detection method together allows the evaluation of the attack strategy and, dually, that of the detection method. Recent incidents highlight the risks of set point manipulation in water infrastructure. In 2025, a reported incident in Norway involved attackers manipulating the reference values of a dam’s control system. At one point, the adversaries attempted to inject an implausible 999% input, which made the attack easily detectable rather than covert. 2 2 2 https://www.csoonline.com/article/4042449/russia-linked-european-attacks-renew-concerns-over-water-cybersecurity.html In November 2023, the Iranian-linked group Cyber Av3ngers compromised the Municipal Water Authority of Aliquippa, targeting a Programmable Logic Controller (PLC) that regulated water pressure and forcing operators into manual control. 3 3 3 https://www.cisa.gov/news-events/alerts/2023/12/01/iranian-cyber-actors-cyberav3ngers-compromise-us-water-utility To assess the resilience of water treatment plants against such threats, this work evaluates both the effectiveness of these covert attack strategies and the ability of detection mechanisms to identify anomalies. The proposed approach builds upon existing methods for covert controller design and attack detection ( 26 ; 3 ; 13 ) , providing a structured approach for analyzing vulnerabilities in networked control systems (NCS). This paper explores change-of-reference attacks ( 9 ) — demonstrating how an adversary can execute them while remaining undetected. The findings contribute to a broader understanding of cybersecurity risks in industrial control systems and highlight the need for stronger defenses against more sophisticated cyber threats. Related Work. While a large body of research has focused on cyberattacks targeting the electrical grid ( 28 ; 10 ) , comparatively fewer studies address water treatment systems ( 29 ) . These facilities often operate in a decentralized manner, with limited staffing, budget constraints, and outdated infrastructure ( 1 ) . As a result, they lack comprehensive cybersecurity measures and remain vulnerable to adversaries capable of launching stealthy attacks. This motivates the need for systematic models and detection strategies tailored to the unique operational and economic characteristics of water treatment infrastructure. The Secure Water Treatment (SWaT) ( 16 ) and Water Distribution (WADI) ( 15 ) datasets are among the most widely used benchmarks for studying cybersecurity in water infrastructure. These datasets provide detailed network traffic and physical sensor data collected from operational water treatment testbeds, and have been instrumental in advancing data-driven attack and detection techniques. Nonetheless, as with any modeling effort, a trade-off must be made between expressivity and simplicity. In this work, we prioritize simplicity to facilitate a principled evaluation of how factors such as model estimation error and measurement noise affect both attack effectiveness and detection sensitivity. Therefore, complementary to the SWaT and WADI efforts, we study a less-explored but still realistic plant model introduced in ( 27 ) . This model offers a simpler analytical structure that enables clearer insights into the dynamics of attack and defense. We apply the covert control scheme from ( 26 ) in combination with the PASAD anomaly detection strategy from ( 3 ) , allowing for a control-theoretic analysis that highlights the interplay between system identification, attacker design, and detection performance. Contributions This paper makes the following three contributions to the study of cyber-physical security in water treatment systems: (1) Attack-Defense Scheme for Covert Control. We develop a structured approach that integrates the covert control architecture of Smith ( 26 ) for man-in-the-middle (MitM) attacks with a detection strategy based on the PASAD anomaly detector. Then, we evaluate how the accuracy of a system identification attack influences the interaction between covert MitM attacks and a detection system. (2) Application to Realistic Water Treatment Models. We instantiate the proposed approach using a linear time-invariant (LTI) model derived in previous work from real-world data collected at an operational water treatment facility. We design and evaluate change-of-reference (CR) stealthy attacks ( 9 ) and show how attackers can exploit system identification to remain undetected. To the best of our knowledge, this is the first work to design covert control-based attacks tailored to water treatment infrastructure. (3) Detection Analysis with PASAD vs. CUSUM. We analyze the detection performance of PASAD under different attack profiles and compare it with the widely used CUSUM detector in process control applications ( 18 ) . Our results demonstrate that PASAD is more