Dynamic Watermarking: Active Defense of Networked Cyber-Physical Systems (1606.08741v1)

Abstract: The coming decades may see the large scale deployment of networked cyber-physical systems to address global needs in areas such as energy, water, healthcare, and transportation. However, as recent events have shown, such systems are vulnerable to cyber attacks. Being safety critical, their disruption or misbehavior can cause economic losses or injuries and loss of life. It is therefore important to secure such networked cyber-physical systems against attacks. In the absence of credible security guarantees, there will be resistance to the proliferation of cyber-physical systems, which are much needed to meet global needs in critical infrastructures and services. This paper addresses the problem of secure control of networked cyber-physical systems. This problem is different from the problem of securing the communication network, since cyber-physical systems at their very essence need sensors and actuators that interface with the physical plant, and malicious agents may tamper with sensors or actuators, as recent attacks have shown. We consider physical plants that are being controlled by multiple actuators and sensors communicating over a network, where some sensors could be "malicious," meaning that they may not report the measurements that they observe. We address a general technique by which the actuators can detect the actions of malicious sensors in the system, and disable closed-loop control based on their information. This technique, called "watermarking," employs the technique of actuators injecting private excitation into the system which will reveal malicious tampering with signals. We show how such an active defense can be used to secure networked systems of sensors and actuators.

• The paper proposes dynamic watermarking as an active defense mechanism that embeds random signals into control commands to detect malicious tampering in CPS.
• It details a methodology using actuator tests and statistical consistency checks tailored for various system types such as SISO, MIMO, ARX, and ARMAX.
• Rigorous simulations demonstrate that significant adversarial distortions in sensor data are statistically detectable, thereby enhancing CPS security and reliability.

Dynamic Watermarking: Active Defense of Networked Cyber-Physical Systems

This paper, authored by Bharadwaj Satchidanandan and P. R. Kumar, addresses the critical need for securing networked cyber-physical systems (CPS), especially given their increased deployment across vital sectors such as energy, healthcare, and transportation. The authors introduce an innovative approach termed Dynamic Watermarking, a technique designed to actively guard CPS against malicious actions by embedding verifiable patterns into system signals.

Problem Formulation and Challenges

Cyber-physical systems integrate physical processes with networked digital control systems, making them susceptible to cyber threats that could lead to severe economic and safety consequences. The fundamental challenge is that unlike purely cyber systems, where security can be enhanced through encryption and secure protocols, CPS involve physical components that require interaction with the environment through sensors and actuators. These interactions present additional vectors for attack, not addressed by traditional network security measures.

The authors differentiate CPS security problems from classical communication network security by noting that in CPS, even authorized individuals can maliciously affect the system, a challenge not entirely addressed by network layer security protocols.

Dynamic Watermarking: Concept and Implementation

The core idea of Dynamic Watermarking involves injecting known, random excitation signals into the system via actuators. These signals serve as a watermark, allowing the system to detect any malicious tampering by sensors. The technique works by superimposing a private, random sequence over the control commands, and then using statistical consistency checks across system outputs to detect deviations indicative of unauthorized manipulations.

• Actuator Tests: The method includes designing tests that involve checking the variance and consistency of the modified control response with the known watermarks. If sensors or actuators deviate from expected behaviors when these random excitations are included, it indicates the presence of malicious activity.
• Adaptability to System Types: The paper demonstrates that this method is applicable across various system types, including single-input-single-output (SISO), multi-input-multi-output (MIMO) systems, auto-regressive (ARX), and more complex ARMAX models. For each, the authors describe how dynamic watermarking can be tailored to accommodate system-specific noise profiles and configurations.

Key Results and Implications

The paper presents rigorous proofs showing that with dynamic watermarking, any malicious attempt to distort or bias sensor data becomes statistically detectable. The authors show that a colluding adversarial node cannot introduce substantial errors without being exposed by the watermarking tests. The effectiveness of the technique fundamentally relies on the adversary being unable to perfectly predict the watermark.

• Numerical Results: Through theoretical simulations, the application of appropriate watermarking significantly constrains an adversary, such that the additional distortion they can introduce to the system remains statistically zero if they are to avoid detection.
• Practical Significance: This technique offers a layer of defense that transforms the problem of cyber-attacks on CPS into one of signal testing. It enables CPS operators to detect and isolate malicious nodes, thus securing the system against unauthorized disruptions.

Extensions and Future Directions

While this paper lays foundational principles for watermarked CPS security, it also opens avenues for further research, including practical system implementations and fine-tuning for different industry applications. One interesting future direction could be extending watermarking techniques to non-linear dynamical systems, thereby broadening security assurances in more complex CPS architectures.

Dynamic Watermarking thus represents a strategic shift in CPS security, aligning with the physical world's dynamics while ensuring the robustness and reliability of these critical systems against sophisticated cyber threats.