Artificial Intelligence and Blockchain for Social Media Security: Authentication, Scam Detection and Federated Learning

Background: Social-media security increasingly involves threats that remain difficult to address through login controls alone, particularly fake accounts, impersonation, phishing, account takeover and behaviour that becomes malicious after successful authentication. Aim: This review examines how Artificial Intelligence (AI), blockchain, sequential behavioural models and Federated Learning (FL) can be combined as complementary security mechanisms for social-media environments. Method: A structured narrative review was undertaken using peer-reviewed journal articles, conference papers, surveys and foundational studies relevant to social-media cybersecurity, decentralized identity, AI-based fraud detection, behavioural modelling and blockchain-enabled FL. Recent literature from 2021–2026 was emphasized, while earlier seminal studies were retained where required to explain RNN/LSTM and FL foundations. The evidence was synthesized by security function, technical approach, reported contribution and limitation. Results: The reviewed studies show a consistent division of strengths. AI is effective for adaptive fraud and behavioural detection; blockchain strengthens credential verification, integrity and accountability; sequential models capture temporal behaviour; and FL reduces the need to centralize raw behavioural data. Recent blockchain-enabled FL studies improve coordination and accountability but continue to face scalability, poisoning, privacy leakage and deployment overhead. The principal gap is the limited empirical evaluation of end-to-end architectures that join wallet-based authentication, post-authentication behavioural analysis, privacy-preserving collaborative learning and blockchain-based accountability.

Download Paper