Gabriele Gualandi: Worst-case Impact assessment of stealth attacks to control systems

Abstract:  

Short description

Hackers could target critical physical infrastructures such as nuclear reactors, electricity grids, or smart cities. If an attack is highly sophisticated,it might remain undetected by anomaly detection systems. We explore an approach to assess the maximum damage that such an attack could inflict.

Long description

Control systems play an important role in modern society. The need for these systems to operate safely and effectively cannot be overstated. To preserve their integrity and functionality, these systems must be capable of self-defense against malicious cyberattacks. Anomaly Detection Systems (ADS) operate by comparing real-time operational data with a predefined model to identify discrepancies, to detect cyberattacks. There is the complication that control systems operates within unpredictable environments. Consequently, ADSs must accommodate minor discrepancies to avoid false alarms that could suggest cyberattacks where there are none. Nonetheless, this tolerance for minor deviations opens control systems to one of the most significant challenges in security: stealth attacks. These attacks subtly manipulate a control system's data, introducing discrepancies that remain below detection thresholds—thus, the attack remains undetected while still compromising the system to some extent. We explore an optimization-based approach to quantify the maximum deviation a stealth attack can introduce into a control system.

Language: ENG

About me:

Following my Master of Science in Artificial Intelligence and Robotics, I pursued a PhD in Computer Science, both degrees from Sapienza University of Rome, Italy. My research focuses on the cybersecurity of control systems. Currently, I hold the position of Senior Lecturer at Mälardalen University in Sweden, where my primary research interest lies in numerical methods for detecting cyberattacks. Additionally, I teach courses in Control Theory and Industrial Robotics.