Engineering & simulation1 to 2% CPU overhead
The company uses a neural network on the edge to analyze time-series data of system calls for key processes, predicting the next expected system call to detect deviations from normal behavior and identify security anomalies.
“We look the time series of CSOs for some key processes … we use this data to develop models that look at like a past window of CSCOs and use that context to predict what the next CSCO is going to be. … if the actual behavior starts diverging from the one it has learned, then we start to see this score increasing.”
Giovanni Alberto Falcione · Hear it at 47:29