More and more AI algorithms are developed for ICT and cyber security applications for private companies and the private sector in and outside the European Union. Often these algorithms function in a non-transparent manner, causing high risks in using them and making legislation challenging. The researchers of the SPATIAL project work on developing and designing resilient accountable metrics, privacy-preserving methods, verification tools and system solutions for a more trustworthy AI.
The researchers are focusing on two specific gaps: the model training of the AI – prio to its use - for bias, privacy and model poisoning and the black box working of the algorithm once it can be used. On one hand they are working on the computational aspects, developing mechanisms that can detect and evaluate faults and set-up tailored measures on how to deal with such circumstances. On the other hand, the researchers in SPATIAL are working on the communication side, providing accountable explanation for the customer or government on how the application works, opening the black box so the users are better able to assess the risks involved.

Based on the findings of SPATIAL, researchers expect there will still be much research needed. The results of the project can be considered a stepping-stone to achieve trustworthy transparent and explainable AI for cybersecurity solutions. The results can also contribute to the governance and regulative framework, setting an international standard for AI driven cyber security to be more secure in the future.