Pole-Dynamics Attacks Detection in Multiagent Systems by Distributed Additive Watermarking.

Area of Science:

• Control Systems Engineering
• Cybersecurity for Multiagent Systems
• Signal Processing

Background:

• Traditional multiagent systems (MASs) lack effective methods for detecting stealthy distributed pole-dynamics attacks (dPDAs).
• Existing single-agent watermarking techniques can affect system states and are insufficient for distributed attack detection.

Purpose of the Study:

• To develop a novel distributed additive watermarking (DAW) approach for detecting dPDAs in MASs.
• To enable the isolation of compromised communication links within MASs under dPDAs.
• To quantify the relationship between DAW parameters and dPDAs detection performance.

Main Methods:

• Implementing DAW by adding and removing watermarks from control signals to avoid system state interference.
• Utilizing distinct watermarking signal covariances for each agent to facilitate compromised link identification.
• Quantifying dPDAs detection performance based on watermarking covariance and developing a DAW-based link isolation scheme.

Main Results:

• The proposed DAW method effectively enables dPDAs detection in MASs.
• A direct proportionality is established between dPDAs detection performance and the sum of watermarking covariances in compromised links.
• The DAW-based scheme accurately isolates compromised links by comparing detection functions and their approximations.

Conclusions:

• DAW is a viable strategy for enhancing the security of MASs against dPDAs.
• The developed method mitigates the adverse impacts of dPDAs on MAS operations.
• Simulation results validate the theoretical findings and the effectiveness of the proposed DAW approach.