Publications by authors named "Cole Sodja"
Technometrics
December 2021
Article Synopsis
- The authors propose a new real-time method for detecting unusual activity in large, sparse networks by using a dynamic logistic model that incorporates various factors influencing sender and receiver behaviors.
- They utilize a variational Bayesian approach to estimate hidden attributes for nodes, allowing these attributes to shift over time to reflect changes in network activity.
- Their algorithm, tested on a network of over 25,000 computers, successfully identifies a red team attack with a detection rate significantly better than previous models without these latent factors.
