Illegal, Unreported, and Unregulated (IUU) fishing represents a critical challenge to sustainable fisheries management and the preservation of marine ecosystems worldwide. The Automatic Identification System (AIS) is a key tool for tracking vessel movements. However, its reliability is often undermined by signal tampering, loss, or disruptions, particularly in remote regions or during adverse weather conditions. To overcome these challenges, this study proposes a novel multimodal framework that combines AIS data with Synthetic Aperture Radar (SAR) imagery to improve the detection of IUU fishing activities. A stacking model was developed, combining multiple machine learning algorithms—Random Forest, XGBoost, and a tailored Multilayer Perceptron (MLP)—to analyze AIS data and detect anomalous fishing patterns. Additionally, Fast R-CNN was employed to process SAR imagery from the Xview3 dataset, enabling accurate vessel identification even when AIS signals were absent. Experimental results reveal that our proposed approach achieves a detection accuracy of 99.75% and a recall rate of 0.998 for drifting longline fishing, markedly surpassing the performance of conventional AIS-based methods. By overcoming the inherent shortcomings of AIS data, this multimodal strategy offers a robust, comprehensive, and effective solution for monitoring and combating IUU fishing on a global scale.