Digital social networks are now becoming an essential part of interpersonal connections and interactions, information exchange, digital identity formation. On the other hand, their rapid expansion has also enabled the widespread creation of fake profiles used for misinformation campaigns, financial fraud, phishing attacks, identity theft, and coordinated influence operations. These accounts are often operated either by automated bots or organized human networks and increasingly function across multiple platforms simultaneously. Most existing detection systems are platform-specific and fail to capture cross-platform behavioural patterns exploited by modern threat actors. This article outlines an extensive review of fraudulent accounts detection across multiple online platforms to overcome hurdles regarding misinformation and cybercrime. Various approaches, including machine learning, deep learning, graph-based, and hybrid strategies are analyzed and compared based on performance, scalability, and adaptability.The study highlights key limitations such as lack of cross-platform generalization, dataset constraints, and limited interpretability. A unified framework is discussed to improve detection efficiency and robustness. The findings provide insights into current trends and future research directions for secure and scalable social media analysis.