The Effects of Ant Colony Optimization on Graph Anonymization

Abstract

The growing need to address privacy concerns when
social network data is released for mining purposes has
recently led to considerable interest in various
techniques for graph anonymization. These techniques
and definitions, although robust are sometimes difficult
to achieve for large social net-works. In this paper, we
look at applying ant colony opti-mization (ACO) to two
known versions of social network anonymization,
namely k-label sequence anonymity, known to be NPhard
for k ≥ 3. We also apply it to the more recent work
of [23] and Label Bag Anonymization. Ants of the artificial
colony are able to generate successively shorter
tours by using information accumulated in the form of
pheromone trails deposited by the edge colonies ant.
Computer simu-lations have indicated that ACO are
capable of generating good solutions for known harder
graph problems.
The contributions of this paper are two fold: we
look to apply ACO to k-label sequence anonymity and
k=label bag based anonymization, and attempt to show
the power of ap-plying ACO techniques to social
network privacy attempts. Furthermore, we look to
build a new novel foundation of study, that although
at its preliminary stages, can lead it ground breaking
results down the road.