Using Coloured Acyclic Nets to Detect Fake Accounts on Social Media

Fake accounts pose a significant threat, not just in spreading misinformation but also in shaping users' perspectives. Beyond this, the establishment of fake accounts serves diverse purposes, spanning from misleading advertising to artificially boosting follower counts. Within the scope of our research, we present a model using Coloured Acyclic Nets (CA-nets). This model is carefully constructed to fulfil the objective of screening, analysing, and identifying fake accounts on social media. Our aim is to enhance the trustworthiness and credibility of online platforms by offering a robust tool for detecting and identifying fake accounts. In this paper, the design of our classifier is based on the MIB dataset [1] consisting of the total of 5301 genuine and fake user accounts. The CA-net model demonstrated respectable performance metrics, yielding solid results in its evaluation. Notably, it achieved the Precision rate of 95%, indicating the model's accuracy in correctly identifying relevant instances, and the overall Accuracy level stands impressively at 97%. Furthermore, the F1 score, a comprehensive measure that balances precision and recall, reached a commendable 96%. These metrics collectively highlight the robustness and reliability of the CA-net model in effectively fulfilling its intended purpose.