Paper presentation at ASIACCS 2018 marks the first success of TU Delft and SUTD collaboration on Cyber Security

Qin Lin had the honour to present, in presence of SUTD professor Aditya Mathur, the paper “TABOR: A Graphical Model Based Approach for Anomaly Detection in Industrial Control Systems” , at the ACM Asia Conference on Computer and Communications Security (ASIACCS) 2018. With an acceptance rate this year of only 20.0%, this marks the first tangible success of the TU Delft – SUTD collaboration, that started less than a year ago. The joint work provides an interpretable machine learning solution for intrusion detection in physical process of industrial control system (ICS). The work was praised as innovative and original work, on a timely and significant subject.

The paper TABOR presents anomaly localization problem in very large dimension data from ICS, i.e., it is able to tell system operators which stages, sensors, or actuators are potentially under attack. Compared with standard “black-box” machine learning techniques, TABOR is more accurate (successfully detect more attacking scenarios with low false positive) and more efficient (only be trained within a couple of minutes). The experiments are conducted in the Secure Water Treatment testbed in SUTD.

The next step is implementing real-time detection version by writing the specification rules learned into the SWaT system. This means such a system is able to read stream data and report the detection results in an on-line fashion.

May many successes follow this great start.