SYSTEM AND PROCEDURES FOR RECOGNIZING A BASIC APPROACH

摘要

A computer-implemented procedure involves training a machine learning system to detect an malicious attack. The procedure involves classifying a first sequence as belonging to a first class indicative of a nominal sequence, based on a first forecast, the first sequence contains an undisturbed version of sensor data. Furthermore, the procedure includes classifying a second sequence as belonging to a second class indicative of an malicious sequence, based on a second prediction,the second sequence contains a disturbed version of the sensor data. Combined loss data are generated for a collection of sequences and are based on a first average loss in relation to wrong first class classifications and a second average loss in relation to wrong classifications of the second grade. The parameters of the machine learning system are updated based on the combined loss data. Once trained, the machine learning system is able to produce a first marking to show that an input sequence is classified as belonging to the first class and to produce a second marking,to show that the input sequence is classified as second class, enabling a control system to operate in a nominal way based on first class and defensively based on second class.