Systematic Review of Artificial Intelligence Techniques Datasets and Weaknesses in Finance Cybersecurity

The persistent challenges posed by increasingly sophisticated financial fraud and cyberattacks underscore an urgent need for adaptive and robust security solutions in digitally transformed financial systems. This study undertakes a comprehensive systematic literature review, rigorously exploring the transformative role of Artificial Intelligence (AI) in bolstering finance and cybersecurity. These primary aim is to meticulously analyze diverse AI techniques, identify the characteristics and limitations of datasets utilized, and delineate inherent weaknesses impacting AI model effectiveness in these critical domains. Employing a stringent PRISMA-guided methodology, 67 high- quality peer-reviewed articles published between 2021 and 2025 were systematically selected from the Scopus database, ensuring a focused and current synthesis. Findings reveal that AI-driven methodologies encompassing advanced machine learning algorithms (e.g., SVM, Random Forest, Neural Networks), deep learning architectures (e.g., LSTM, GRU, DNN), and privacy-preserving approaches like federated learning significantly enhance fraud detection, anomaly detection, and overall data security. Noteworthy strengths include high detection accuracy, real-time monitoring capabilities, and operational scalability. However, critical weaknesses persist, primarily stemming from heavy dependence on high-quality, representative datasets, significant computational demands for real-time processing, and complex integration requirements with legacy systems. This review concludes that despite these limitations, AI holds immense potential for fostering trust, resilience, and innovation within financial ecosystems. It critically emphasizes the imperative for developing robust data strategies, ensuring ethical AI deployment frameworks, and maintaining rigorous methodological refinement to fully harness AI’s impact on securing and advancing global financial and cybersecurity systems.