SPATIAL Project encompasses various use cases, including the Resilient Cybersecurity Analytics use case. WithSecure, the leading organization in this domain, studies evasion and poisoning attacks against ML models used in cybersecurity while exploring effective defences against such threats.
Given the scarcity of online trained dynamic attack detection systems provided by third parties for experimentation, SPATIAL adopts a proactive approach. Research prototypes are developed as a pilot, enabling the implementation and evaluation of novel techniques and methodologies.
In the dynamic detection of advanced attacks, a prevailing approach entails the deployment of sensors within the protected environment. These sensors capture relevant data streamed to a dedicated security backend. The data is thoroughly analysed through a combination of machine learning methods, detection rules, and human expert analysis. Subsequently, informed decisions are made to mitigate and respond to potential threats.
This comprehensive approach allows SPATIAL to explore new avenues for bolstering cybersecurity, while the Resilient Cybersecurity Analytics use case specifically focuses on fortifying ML models against sophisticated attacks. WithSecure’s involvement ensures that state-of-the-art defences are devised and implemented, pushing the boundaries of cybersecurity resilience.
Learn more by watching Samuel Marchal‘s interview! (Samuel is a Senior Data Scientist / Squad Lead (Machine Learning Security) at WithSecure).