Distributed Maximum Correntropy Filtering for Stochastic Nonlinear Systems Under Deception Attacks.

Area of Science:

• Control Systems Engineering
• Signal Processing
• Cybersecurity

Background:

• Stochastic nonlinear systems are vulnerable to deception attacks, compromising system performance.
• Traditional filtering methods often struggle with non-Gaussian noise and distributed computation requirements.

Purpose of the Study:

• To develop a distributed filtering scheme for general stochastic nonlinear systems facing deception attacks.
• To enhance filter robustness and accuracy using the maximum correntropy criterion.

Main Methods:

• Utilized Taylor series expansion and fixed-point iteration for filter gain and error covariance calculations.
• Employed the weighted maximum correntropy criterion, replacing traditional minimum covariance methods.
• Developed a simplified filter by removing weights from the correntropy criterion.

Main Results:

• Derived calculation formulas for filter gains and the upper bound of filter error covariance.
• The derived upper bound relies solely on local, neighbor, and attack statistics, enabling distributed computation.
• Demonstrated the effectiveness of the distributed maximum correntropy filtering through an illustrative example.

Conclusions:

• The proposed distributed maximum correntropy filtering effectively addresses deception attacks in stochastic nonlinear systems.
• The method offers a robust and computationally feasible approach for distributed filtering under adversarial conditions.
• The reliance on local information enhances practical applicability in networked systems.