Abstract:
This research had the objective to develop framework and evaluate the chosen methods by extracting attributes from Twitter's data and identifying attributes that are helpful in account cloning detection. Twitter is chosen as a case study due to it being a popular social network platform in Thailand. The framework consists of 3 parts as follows: 1) Twitter crawler retrieved data and their recent tweets from Twitter, 2) Attribute extractor extracted attributes for further analysis, and 3) Cloning detector analyzed the similarities of user profiles, similarities of friend & follower and classification of post behavior & writing style as fake or authentic accounts. The profile attribute similarity, friend & follower similarity can help identify accounts that may be attackers. Potential attackers and victim can be put under close monitoring by further experiment. Posting behavior & writing style can further distinguish whether any post claimed to be written by the victims is authentic. Status related attributes are more discriminative than writing style attributes. If using these attributes altogether yields the best results. The experimental results support that the proposed framework is feasible and effective. It can be customized to support other social network platforms.