Fast efficient evaluation of messages on automotive networks using look-up tables

A non-transitory storage medium storing instructions readable and executable by a microprocessor or microcontroller of an Electronic Control Unit (ECU) of a vehicle to perform a classification method for classifying Controller Area Network (CAN) messages, the classification method comprising:

• receiving CAN messages from a CAN bus of the vehicle via a transceiver of the ECU;
• for each received message, applying a Support Vector Machine (SVM) classifier employing a kernel ƒ(xi,y)∝Πd=1DF(xi,d,yd) or a kernel ƒ(xi,y)∝Σd=1DF(xi,d,yd) to classify a vector y representing the received CAN message, where D is the dimensionality of Support Vectors xi;
• and
• causing the ECU to output an alert if the classifications of messages by the SVM classifier meet an alert criterion;
• wherein the applying includes determining at least some terms F(xi,d,yd) of the kernel ƒ(xi, y) using a look-up table.

A method comprising:

• using a server computer, generating a Support Vector Machine (SVM) classifier by operations including:
• (1) training an SVM classifier employing a kernel ƒ(xi,y)∝Πd=1DF(xi,d,yd) or a kernel ƒ(xi,y)∝Σd=1DF(xi,d,yd) where D is the dimensionality of Support Vectors xi and of data vectors y;
• and
• (2) generating a look-up table containing the possible values for at least one term F(xi,d,yd) of the kernel ƒ(xi, y) which term F(xi,d, yd) has a finite number of possible values;
• and
• using a microprocessor or microcontroller of an Electronic Control Unit (ECU) of a vehicle having a Controller Area Network (CAN) bus:
• (3) receiving messages from the CAN bus via a transceiver of the ECU and, for each message, applying the trained SVM classifier to classify a vector y representing the received message, wherein the applying includes determining at least some terms F(xi,d,yd) of the kernel ƒ(xi, y) using the generated look-up table.