European Patent Office to Grant Clavister’s Patent Application for AI
Örnsköldsvik, Sweden - 12 November 2024 - Clavister, a leader in European cyber security for mission-critical applications, today announced that the European Patent Office (EPO) has issued an “Intention to Grant” for Clavister’s patent application covering the company’s PASAD AI technology.
The patent will provide Clavister’s PASAD (Process Aware Stealthy Attack Detection) AI technology with protection in a number of European countries. The “Intention to Grant” status means that only a formal process remains until the selected countries have approved the patent.
“This is an important and exciting decision by the EPO, and we view this announcement as recognition of Clavister’s efforts to be in the forefront of innovative cyber security technology” said John Vestberg, CEO of Clavister. “This patent provides protection for our PASAD AI technology which has proven to be highly effective in detecting modern day cyber attacks.”
Clavister PASAD is an AI-based anomaly detection technology that monitors data streams in a system and continuously reports the behavioural state of the monitored system. In general, PASAD helps to classify between normal and abnormal system behaviour. This capability is highly relevant in modern cyber threat mitigation, where traditional protection methods fall short. Examples include zero-day cyber attacks where an intruder manages to compromise a part of the system using attack vectors unknown to the public.
As an AI engine, PASAD needs to be trained on data representing the normal behaviour of the system to be monitored. The amount of data needed for training is very small for an AI-based solution, as it requires just enough data to capture the system’s cyclical behaviour. While it varies from system to system, typically a couple of hours worth of data is sufficient. Since PASAD only trains on normal data, it can detect anything that deviates from normal behaviour, including zero-day attacks, as well as anomalies resulting from wear and tear of systems. Both training and inference can be performed locally, so there is no need for a connection to any central or cloud infrastructure to carry out the work.
PASAD is high-performing and resource efficient, with its inference engine processing up to 10 million records per second on a single CPU core. It can further be adapted to many different environments and works for data transmitted over any protocol and media, including IP, CAN, Modbus, fiber optics, satellite radio and so forth.