SOS: Save Our Social Network Accounts
Rana Mohamed Eisa, Merna Labib, Amr El Mougy · 2019
Social media platforms such as Facebook and Twitter have become universally prevalent for connecting people, sharing information, or spreading messages and ideas. However, the social dominance of these platforms have also made them attractive targets for malicious intruders. Successful attacks against personal social media accounts can cause serious harm to their users, due to the possibility of revealing intimate private information. Discovering these hacking attempts can be done by detecting anomalous (out of the normal patterns) behavior over the user's account. However, personal accounts show inconsistent behavioral patterns over time, making it highly challenging to extract definitive features for each user, and accordingly to detect anomalies. In this paper, we propose a novel methodology based on subjective logic and machine learning techniques to detect and extract the most highly defining features of each personal account. These features are then used to build a behavioral profile for each user, which can be used to detect anomalies. We conducted 2 studies, collected data from 47 Facebook users and 616 Twitter users. We extracted features from the collected data and assigned each feature a weight as a scale of importance according to the user's deviation in this feature. We simulated 2,209 different attacks in the first study and 379,456 different attacks in the second study. Our results show a detection accuracy of 94.9% and 87.54% for the first and second studies, respectively.