Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{205708,
        author = {Taniya Nath and Madhumitha H Athreya and Kushagra Maheshwari},
        title = {Machine Learning and Deep Learning Approaches for Income Tax Fraud Detection: A Comparative Review},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {8081-8095},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=205708},
        abstract = {Income tax fraud poses a persistent and evolving challenge for revenue authorities worldwide, with traditional rule-based detection systems increasingly unable to keep pace with sophisticated evasion techniques. This paper presents a comparative review of machine learning (ML) and deep learning (DL) approaches applied to income tax fraud detection, examining their methodologies, performance characteristics, and practical limitations. The review surveys classification techniques including decision trees, random forests, gradient boosting methods, and neural network-based anomaly detection models, evaluating their reported effectiveness across varying data conditions. Because peer-reviewed, income-tax-specific literature remains comparatively limited, the review further extends, under explicit methodological bounds, to transferable techniques from the larger adjacent literature on financial fraud detection, including graph neural networks for relational fraud modelling and federated learning for privacy-preserving cross-institution training, alongside a methodological case study on evaluation flaws such as data leakage that inflate reported performance in credit-card fraud research. Particular attention is given to challenges such as class imbalance, limited access to public tax datasets, and the trade-off between model accuracy and interpretability, a critical concern for regulatory and compliance applications. The paper concludes by identifying gaps in current research and proposing directions for future work, including hybrid models, explainable AI frameworks, and India-specific opportunities to apply graph-based and federated techniques to the existing GSTN and CBDT analytics infrastructure. This review aims to provide researchers, policymakers, and financial institutions with a consolidated understanding of the current ML/DL landscape in tax fraud detection.},
        keywords = {Anomaly Detection, Explainable AI, Federated Learning, Fraud Detection, Graph Neural Networks, Income Tax, Machine Learning, Tax Compliance.},
        month = {June},
        }