Characterization of Latent Social Networks Discovered through Computer Network Logs

The collection of true and complete measurements for social network analysis is often limited and costly. Indirect measurements from electronic data sources have opened new avenues for extracting valuable information on social patterns. While the analysis of proximity data and email communication are a few data sources that have been heavily studied, the analysis of computer network logs and their robustness in capturing latent social behavior is an under-explored area. In this paper, we investigate the richness of network logs in identifying latent social networks of users within an enterprise. Specifically, we model the observable network artifacts on a per-user basis, and embed each user in graphs representing the enterprise. We analyze the structural properties of these graphs and discover that the latent social network shares many of the distinguishing characteristics of those observed in the real-world.