Security posture assessment of power information networks based on adversarial deep learning and D-S evidence theory

To address the shortcomings of over-reliance on expert experience in the construction of the basic probability assignment of traditional D-S evidence theory, a security posture assessment model for power information networks against deep learning and D-S evidence theory is proposed. Firstly, sample data are collected from different security detection devices and augmented using a generative adversarial network. Secondly, a genetic algorithm is used to optimize the back propagation neural network to obtain the basic probability assignments of different evidences. Finally, the D-S evidence theory is used to synthesize and calculate the trust degree matrices of different posture states according to the temporal order to successfully complete the posture assessment task. The proposed evaluation method is experimentally verified to be feasible and effective.