• A Boeing 777 from Perth to Kuala Lumpur experienced a severe failure which forced the pilot to disconnect the autopilot and return to Perth after the speed decreased from 270 kts to 158 kts.
• While in cruise at 37,000 ft, in response to spikes in angle of attack (AOA) data, the Airbus A330 flight control primary computers (FCPCs) commanded the aircraft to pitch down. At least 110 of the 303 passengers and nine of the 12 crew members were injured.

Different approaches are possible to solve this issue. On-line testing uses routines, performed at runtime, to test the behavior of the system when in idle state. This mainly aims at identifying defects. Sometimes that is not a strong enough guarantee if the system is performing safety-critical operations. For such situations, online monitors continuously analyze the behavior of a system while it is operating.
A monitor observes the behavior of a system and detects if it is consistent with a given specification. In more detail, the monitor observes particular signals and, if some incorrect behavior or an anomaly is detected, raises a warning. Such an approach is currently used in vastly different domains: from intrusion detection systems (IDS) in distributed network applications to single IoT devices. There are various paradigms for monitors using data-driven approaches, machine learning, concurrent models of the system, or automated generation from the specification.
This seminar will give a prospective on the state of the art of monitoring techniques.
Each participant starts from an original research paper, prepares a presentation including a demonstration or example, and submits a report.

Language: Deutsch/ English