<p>Detecting fraudulent trading activities in the FinTech sector presents significant challenges due to the highly dynamic nature of financial markets and the rapid adaptation strategies employed by fraudsters. Conventional machine learning techniques, especially when trained on imbalanced datasets, often exhibit classification bias toward the majority class (legitimate transactions), reducing their ability to identify fraudulent behaviors. To address this limitation, we introduce a hybrid memetic-ABC framework that combines Artificial Bee Colony (ABC) sampling for global exploration with memetic local refinement to generate high-quality synthetic minority-class samples guided by feature importance. ABC ensures broad coverage of potential fraudulent patterns, while the memetic refinement fine-tunes synthetic samples and feature subsets to enhance discriminative power, capturing subtle anomalies in transaction amounts, locations, and timing. The framework further incorporates an anomaly detection module that constructs behavioral profiles for traders using historical transaction data, clustering them based on similar trading patterns. This enables identification of key indicators of potential fraud, such as sudden volume fluctuations, irregular transaction timings, and deviations from behavioral norms. For real-time monitoring and scalable deployment in decentralized environments, the system is extended into a Distributed Intrusion Detection System (DIDS) architecture, simultaneously detecting transactional anomalies and network-level threats, including DDoS attacks and unauthorized access attempts. By correlating anomalous trading patterns with DIDS-generated alerts, the approach achieves enhanced detection accuracy and robustness against complex fraud scenarios. Experimental evaluations demonstrate that the proposed model outperforms state-of-the-art techniques, achieving approximately 10% improvement in accuracy.</p>