This is a living document for the ROS 2.0 "Intrusion" Detection System development effort supported by TARDEC in Warren, MI and Carnegie Mellon's SEI Emerging Technology Center in Pittsburgh, PA.
This project is a 6.1 research project to attempt to create a holistic anomaly detection in ROS 2.0 that can detect and alert the operator of node failures, system degradation, faulty system states, irregular behavior, and more. The ideal IDS would monitor all systems, nodes, and services on the ROS 2.0 platform and perform detection, real-time learning, and alerts accordingly. One of the main goals of this project is to see what ROS 2.0 is lacking/limiting in in order to guide development of ROS 2.0.
# Current Progress
# Point(s) of Contact
1. Ritwik Gupta - Carnegie Mellon (rgupta at sei dot cmu dot edu)
I've done a research project of my own building a learning-based anomaly detector for industrial ROS applications. We have called it the ROS ADM (Anomaly Detector Module). Feel free to give it a try. I've created the package, but am still in the process of filling in the README. Here is the [repo](https://github.com/narayave/ros_anomaly_detector).
Just a couple thoughts on it. While it was built for industrial applications, I think it can be extrapolated to other ROS applications. This should be the focus moving forward. There are some limitations to the ADM, and I'll address them in the README.
It was built for ROS 1, but possible to re-build it for ROS 2. I am continuing to gain understanding of ROS 2, and I'm hoping the ADM project can help with the formulation of an IDS.