This talk will explore learning-based testing techniques, with a focus on automata learning, that can verify robustness and timing requirements of complex software systems. Traditional testing approaches can be inadequate in detecting subtle errors and corner cases, leading to unexpected system behavior and failures. By leveraging machine learning and automata-based techniques, high-quality test cases can be generated, subtle bugs can be detected, and timing constraints can be verified. We we will discuss our research and present case studies in fuzzing and model learning of IOT protocols (Bluetooth LE, MQTT), automotive controllers, and web-services.

Bernhard K. Aichernig is a tenured associate professor (ao. Univ.-Prof.) at Graz University of Technology, Austria. With his research group he investigates the foundations of software engineering for realising dependable computer-based systems. Bernhard is an expert in formal methods and testing. His research covers a variety of areas combining falsification, verification, and abstraction techniques. Current topics include the Internet of Things, test-case generation, automata learning, and statistical model checking. Since 2020, he leads the TU Graz - SAL Dependable Embedded Systems Lab (DES Lab) with fundamental research in zero-bug software and dependable AI. He participated in five European projects. From 2004-2016 Bernhard served as a board member of Formal Methods Europe, the association that organises the Formal Methods symposia. From 2002 to 2006 he had a faculty position at the United Nations University in Macao S.A.R., China. Bernhard holds a habilitation in Practical Computer Science and Formal Methods, a doctorate, and a diploma engineer degree from Graz University of Technology.