The Foreign Exchange (Forex) market is a decentralized, high-frequency trading environment that is particularly vulnerable to manipulation and fraud. Fraudulent strategies often exploit low-liquidity periods to distort price movements and mislead traders. While traditional fraud detection strategies are simple and interpretable, their effectiveness and efficiency are constrained in real industrial scenarios due to the need for extensive expert verification and significant time investment, making it challenging to identify complex fraud patterns. To tackle these challenges, we propose a Forex market fraud detection (FMFD) framework named Accelerated Ensemble Fraud Analysis (AEFA). It consists of three key components: an Initial Data Optimization module for outlier removal, a Probabilistic Decision Module employing Random Forest for high-recall classification, and an Advanced Voting Integration module with a soft voting strategy. Extensive experiments demonstrate that AEFA outperforms existing methods with superior performance and efficiency, enabling scalable fraud detection in Forex trading systems.