QuantPi: Subgroup Discovery to determine Time of Applicability of malware detection systems
QuantPi is one of five winners of the first ELSA Industry Call and has received 60,000 euros in funding, working together with top AI experts within our Network of Excellence for six months. QuantPi helps organizations safely accelerate their AI transformation through aligning AI lifecycle stakeholders around scalable AI testing on relevant performance, risk and compliance dimensions.
In this video, QuantPi presents the joint project with ELSA:
QuantPi’s short summary of the project:
“The number of internet-enabled devices we use, is ever increasing and therewith so does the attack surface and incentive for malicious actors to infiltrate these devices, e.g. via malware applications. To counteract, many malware detectors exist. Due to the arms race between malware developers and malware detectors, the malware landscape changes both rapidly and suddenly over time. This results in increasing error rates necessitating frequent inspection and reevaluation of detectors by human experts. To increase the reliability of malware detectors, and potentially automate adjustments in the wake of unforeseeable data drifts, it is therefore highly relevant to be able to detect and quantify this data drift.
In our collaboration with ELSA and CISPA, we showed that recent subgroup discovery techniques are powerful tools enabling such data drift quantification while providing interpretable associations to prediction errors.”