Tom Mahler from Ben-Gurion University has shared a solution that will protect medial medical devices such as CT, MRI, Ultra sound systems that use a host control PC to send instructions to the gantry by developing two layered architecture: (1)an unsupervised context-free layer that detects anomalies based solely on the instructions' content and intercorrelations; and (2) a supervised context-sensitive layer that detects anomalies with respect to the classifier's output, relative to the clinical objectives. This solution improved anomaly detection performance from 71.6% to 82.3%-98.8%.
Why do we need this solution in addition to Firewalls and other network monitoring solutions? Number one, there are big gaps in the context of these devices are being used. The same device is used for kids and adults and the device does not know which patient is being evaluated which can introduce potentially harmful threats to patients such as radiation overexposure. Number two, the host is closed system, so security updates cannot be installed to the endpoint devices. This solution can monitor the traffic between the host system and the gantry by plugging in one side and the output goes to the gantry without changing the hosting PC's networks or interface. If the system fails, simply you need to unplug the solution and everything operates without interruption. Lastly, the layer 2 utilizes AI technology to learn its context to detect the anomaly of instruction which helps manufacture to detect the cyber attacks, human errors, or host PC software bugs.
This additional context sensitive layer enables to detect anomalies which cannot be detected using only context free layer that anomalies based solely on the instructions' content and intercorrelations.
https://thecyberwire.com/podcasts/research-saturday/157/notes
No comments:
Post a Comment