Abstract:
Popularized with Milgram's 1967 chain letter experiment, the field of social networks has expanded to be not only a field of academic study, but an invaluable end-user tool. Concurrently, the early experiments in this field were limited in scale by computational constraints and the cost of acquiring the data.
The advent of the world wide web and cheap storage allows for large scale experiments and marks a shift from qualitative approaches to algorithmic ones. Beyond graph analysis and the efficient computation of graph metrics, we are now looking at extracting high level knowledge from the social networks. This includes detecting meetings, cross-checking data quality and exploiting groups for online recommender systems.
The talk closes with the ongoing problem of friend finding and privacy protection within online social networks. In current interfaces, browsing must be used for friends to find each other within the network. This makes the network vulnerable to spam or outright copying. We propose here a decentralized method of friend discovery that re-uses the existing network in a secure fashion.
