D4R: a new direct discrete dynamic data reconciliation method for the detection of cyber attacks

Abstract A novel hybrid method of data reconciliation and gross error detection, applicable for systems with a mixture of dynamic and static system constraints, is developed for the detection of cyber attacks. The requirements for the new application of data reconciliation and similar methods in cybersecurity differ from the requirements for the established use of data reconciliation in automation and control engineering. For the detection of cyber attacks aiming at physical damage the main focus is on significant gross error detection while for classical applications a robust optimization and smoothing of measurement data is the main concern. Therefore the new hybrid method of direct discrete dynamic data reconciliation, as well as similar methods of data reconciliation and Kalman filters with their referring methods of gross error detection are evaluated regarding their aptitude for attack detection in cybersecurity. All considered methods are compared regarding properties resulting from the specific optimization procedure and the detection. The new direct discrete dynamic data reconciliation is indeed shown to outperform the other methods regarding the detection of cyber attacks.