AI research at the front line

Under the direction of the German Federal Ministry of Education and Research adesso and the managed service provider adesso as a Service (3as) support a research project to test solutions for securing modern web infrastructures using AI.

Modern Web services are the basis and driver of the ongoing digitization of our society. They also serve as a trusted source and administrator of digital identities for millions of users. The misuse of these identities must be securely ruled out in order to avert damage to private users and economic players. AI-based detection methods for attacks against IT systems already exist. However, these usually focus on clearly defined characteristics. Attacks against web services, however, often do not only target an isolated technical level. They usually develop over time and leave traces at various points in the system, which must be evaluated holistically. This is where the research project "Artificial Intelligence for secure web infrastructures with digital identity management" (AIWI) comes in.

AIWI is intended to contribute to the development and practical testing of AI-based cognitive security management in complex web infrastructures. The focus is on the development and testing of AI procedures and systems for detecting and fighting possible attacks against web identity services.

AIWI's partners include 1&1 Mail & Media Development & Technology GmbH, Karlsruhe Institute of Technology (KIT), secuvera GmbH, the Karlsruhe University of Applied Science and the Technical University Carolo-Wilhelmina in Braunschweig. adesso SE and adesso as a service GmbH contribute their know-how about AI processes.

The partners are investigating how data from distributed detectors can be merged in order to deduce the existence of attacks from the overall picture. The procedures are to work based on domain-specific and cross-domain collected data or observations and thereby meet the requirements of autonomy, confidentiality, data protection and self-determination of users and operators. The approach of merging the data should therefore prevent the possibility of drawing conclusions about the participating companies based on the data.

The participants are also exploring how they can train AI models based on data collected in different systems. The applicability of AI for securing web infrastructures depends on the quality of the training data. AIWI will develop a framework to meet industrial quality requirements for data management by means of a rigid "data governance".