Distributed Estimation With Cross-Verification Under False Data-Injection Attacks.

Area of Science:

• Distributed Systems
• Cybersecurity
• Signal Processing

Background:

• False data-injection (FDI) attacks compromise distributed algorithms by tampering with agent data, hindering accurate parameter estimation.
• Existing FDI attack detection algorithms often suffer from poor detection rates and low efficiency due to the concealed nature of malicious data.

Purpose of the Study:

• To propose a novel distributed diffusion least-mean-square (DLMS) algorithm integrated with a cross-verification (CV) mechanism to counter FDI attacks.
• To enhance the detection performance and efficiency of distributed algorithms against sophisticated data tampering methods.

Main Methods:

• Introduction of a distributed diffusion least-mean-square with cross-verification (DLMS-CV) algorithm comprising detection and secure estimation subsystems.
• Implementation of a smoothness strategy within the CV mechanism to improve detection capabilities.
• Analysis of algorithm convergence and formulation of an adaptive threshold design.

Main Results:

• The proposed DLMS-CV algorithm demonstrates improved detection performance against FDI attacks.
• Probabilities of missing and false alarms decay exponentially with a sufficiently small step size.
• Simulation experiments validate the effectiveness and simplicity of the DLMS-CV algorithm compared to existing methods.

Conclusions:

• The DLMS-CV algorithm offers a robust solution for detecting FDI attacks in distributed systems.
• The introduced smoothness strategy and adaptive threshold enhance detection accuracy and reliability.
• The algorithm provides a significant advancement in securing distributed estimations against data injection threats.