This in general is a hard problem, but I recommend partnering with a University that has a strong info security and computer science program.
Crowdstrike has a very good system for detecting these "abnormal behavior" events at scale.
You will need to ingest a large amount of data, to a central log server that collects all of the MacOS logs.
With a machine learning approach, you will need a large list of anomalies
You can also try things like an "artificial ignorance" approach, where you alert on the first time a user runs a new program. But it will be hard to filter out
"bad software install" compared with regular installs.
Topic:
Machine Learning & AI
SubTopic:
General
Tags: