Cyber & IT Supervisory Forum - Additional Resources

ARTIFICIAL INTELLIGENCE AND CYBERSECURITY RESEARCH

• Securing the software packages that were used and verifying the validity of the training data 80 . • Approaches addressing adversarial attacks 81 82 that in general are ad-hoc and focused on a specific type of attack assumed to be known a-priori. This is due to the size of the adversarial attack generation space which is potentially of large dimensions. As such, both traditional and neural network-based ML approaches can be used depending on the specifications of the problem-at hand.

80 D. Gümüşbaş, T. Yıldırım, A. Genovese and F. Scotti, A Comprehensive Survey of Databases and Deep Learning Methods for Cybersecurity and Intrusion Detection Systems, in IEEE Systems Journal, vol. 15, no. 2, pp. 1717-1731, June 2021, DOI: 10.1109/JSYST.2020.2992966. 81 Idem footnote 79 82 Yunfei Song, Tian Liu, Tongquan Wei, Xiangfeng Wang, Zhe Tao, and Mingsong Chen. Fda3: Federated defense against adversarial attacks for cloud-based IoT applications. IEEE Transactions on Industrial Informatics, ages 1–1, 2020. DOI:10.1109/TII.2020.3005969

25