The cybersecurity industry is one of the pioneers in the application of artificial intelligence. It has been used for years in attack detection and virus scanners. However, unanswered questions regarding security, data protection, and liability risks keep many companies and authorities from using AI in practice.
It is often difficult for user organizations to assess whether AI systems are even mature enough for practical use. At the same time, manufacturers find it difficult to prove the safety and quality of their AI products. In laboratory environments, AI systems can sometimes recognize attack patterns very well, but not in real Internet environments. Test points and corresponding benchmarks can provide user organizations and manufacturing companies with reliable orientation. Specific minimum requirements from industry associations can provide orientation and the public sector could maintain test centers that can check compliance with the criteria, for example for use in critical infrastructures.
Experts from science and industry worked out recommendations on how obstacles to the use of AI can be overcome: These include specific quality criteria and testing options. The complete results report with all recommendations, many explanations, and practical examples can be downloaded free of charge.
The use of AI systems by attackers and saboteurs poses a particular challenge. Today, companies already use AI systems to protect the company network in order to identify and thwart attacks more quickly. Attackers manipulate the training data of such systems or use AI systems themselves as attack weapons. In this way, attackers can attack a company repeatedly and train the system particularly well. If the attacker changes the target and then attacks another company, the trained system can be much more sophisticated. In order to counteract the use of AI as a weapon of attack.
“In cybersecurity, AI can make a substantial contribution to reacting correctly to cutting-edge and future threats in commercial enterprise and society. A prerequisite for this is that the structures and their offerings continue to be technically assessable and unique minimal necessities are developed for distinctive industries and fields of application”, says Experts.