- Status: Project phase
Background
IT forensics describes the securing and analysis of criminally relevant data from IT systems using scientific methods. Strictly methodical data analysis is used to prove facts in court. Digital forensics also plays a decisive role in the application context of artificial intelligence, as AI systems are often used in a networked manner in companies in addition to their private use. Understanding when and why evolving and self-overwriting AI-based techniques fail must therefore be analyzed and improved using forensic methods and tools.
Aim
The research program “Forensics of Intelligent Systems” addresses the investigation and exploration of existing or fundamentally new forensic approaches, tools taking into account current legal standards. The aim is to provide proactive “forensic readiness” and “forensic-by-design” methods for the legal proof of manipulation in evolving, self-overwriting AI systems. These novel and method-oriented research approaches for the legally compliant proof of misconduct must be transparent, fair and explainable. The methods are to be evaluated and improved using inversion and evasion attacks as well as data poisoning attacks on an AI system. The main topics addressed within the project include image processing, text processing and autonomous systems, with a focus on autonomous driving and its sensor technology.
Disruptive Risk Research
The area of application of IT forensics in AI has no or hardly any comparable research projects. The project is highly relevant to the rule of law, society and the economy, while at the same time posing a high technical and scientific risk. If successful, Germany will assume a leading position in the global economy and society.