Cyber Security Webinar by Dr. Mark Roeling: Unsupervised clustering on spatio temporal data

In Cyberspace, data frequently have a spatial as well as temporal dimension, for example when we capture the activity from a network of computers. Straightforward clustering methods do not always apply in this setting since computer networks are often disassortative, meaning some nodes in the network display similar behaviour but are not in the same cluster.

In this presentation I will discuss our work on unsupervised clustering of Wireshark captures of botnet infected computer networks with a method that clusters edges as well as nodes. We combine a method from spectral clustering with stochastic blockmodels to achieve competitive classification performance. We compare our results with clustering solutions from other well-known methods and replicate our main finding in another dataset.

Zoom Details:
https://tudelft.zoom.us/j/97057185992?pwd=K2p6WFdLYTJ6RjVFdlU0eFQzamk3QT09

Meeting ID: 970 5718 5992
Passcode: 217633