Investigation of Social Media From Content and Structural Perspectives

Investigating Online Social Networks (OSNs) is crucial due to their pervasive influence on modern society, serving as primary platforms for communication, information dissemination, and social interaction. With billions of users worldwide, OSNs play a pivotal role in shaping public discourse, spreading news, and fostering interpersonal connections. Therefore, understanding the complexities of OSNs is essential for addressing emerging challenges, such as misinformation, privacy concerns, and network resilience, in order to contribute to the development of more secure, trustworthy, and resilient OSNs. In this dissertation, we address critical challenges within OSNs from both content and structural perspectives. In the context of content, we focus on how information is changing and being shared, with particular attention on understanding and detecting fake news. Leveraging advanced machine learning techniques, we develop an advanced fake news detection mechanism, integrating both textual and contextual features of the news article for enhanced detection accuracy.

Furthermore, we investigate weaknesses in audio deepfake detection systems, uncovering their vulnerability to adversarial attacks. We propose novel countermeasures to enhance the robustness of deepfake classifiers against malicious manipulation of audio content, thus ensuring the integrity of multimedia circulated across OSNs. From a structural perspective, the limitations of traditional mutual exclusion algorithms in distributed systems, e.g., OSNs, are addressed. A novel failure-resilient token-based mutual exclusion algorithm is proposed, offering a scalable and efficient solution for coordinating access to shared resources. Leveraging a tree structure, the proposed algorithm ensures the reliability and resilience of distributed systems while minimizing communication costs.