Cyber & IT Supervisory Forum - Additional Resources

ARTIFICIAL INTELLIGENCE AND CYBERSECURITY RESEARCH

into how this software can enhance deception clearly needs to be undertaken urgently. Then, hopefully, the training of specialists can become more consistent, taking this development into account.

Training models for practitioners in real-world scenarios

Type: AI training Description:

Develop training on AI for practitioners using real-world scenarios.

Objectives:

1. Training on AI in real-world scenarios

Entities:

• Security researchers • Universities

Beneficiaries:

• Cybersecurity practitioners

Existing research:

Cybersecurity is a never-ending task that is not only about stopping threats from intruders, but also about not wasting time and energy on false positives, which requires a ‘rock-solid' belief in the AI model coupled with rapid escalation to human analysts. The best AI requires data scientists, statistics and as much human input as possible. The foundation for effective 'triage' activity against the multitude of risks and forms of attack is teaching AI when incidents occur, teaching an AI threat disposition system, training practitioners using real-world scenarios and conducting real behavioural threat analysis. This is also what the ‘human in the loop’ 122 interaction requires.

AI & cybersecurity threats Observatory

Type: AI for cybersecurity Description:

122 Jones, Tim. 2019. IBM Developer ‘Take a Look at AI and Security and Explore the Use of Machine Learning Algorithms in Threat Detection and Management. (blog). 19 August 2019. https://developer.ibm.com/articles/ai-and-security/

36