Home » Project of the week » CLAS - Cyber-safe learning control system

CLAS - Cyber-safe learning control system

Date: 
01/01/2018 to 31/12/2023

The advances in ICT have led to a dramatic increase in the number of deployed devices that can communicate, sense, and actuate. Such devices and adjoining computing infrastructures enable a plethora of new applications, but they rely on open systems and networks, which suffer from potential vulnerabilities.

In CLAS, we focus on analyzing and defending the cyber-security of physical infrastructures controlled and monitored using machine-learning algorithms. Machine learning has gained immense popularity, and is being introduced in the real-time operation of large-scale infrastructures. Nevertheless, such algorithms can be sensitive to small but malicious manipulations of data.

To date, no principled, robust design methodology for machine-learning algorithms operating in closed loops exists, and consequently the main scientific contribution of CLAS will be to develop and demonstrate such techniques. The project consists of six work packages covering threat models, robust learning, resilient control, risk allocation, control forensics, and demonstration.

As a demonstrator, CLAS will use the KTH Live-In Lab, incorporating 200 apartments, as a proxy for a smart city. The CLAS consortium consists of five complementary groups at KTH. Individually, all groups can be claimed to be leading in their respective fields, and together they form a world-class research environment, facilitating true breakthroughs in research supporting future cyber-secure learning and control systems.

Week: 
Monday, 3 September, 2018

News

On the event of the adoption of the draft regulation laying down measures for a high common level of cybersecurity at the institutions, bodies, offices and agencies of the Union, the AI4HealthSec project kicked off a process to provide its opinion.