ARCHIVES

Counterfeit currency continues to be a threat to economical stability of a nation and its retail ecosystem. The increasing availability of high-resolution printers and scanners enables forgers to produce banknotes that are closely resembles with the genuine ones, which poses challenging situation to the effectiveness of both manual inspection and legacy sensor-based detectors. In the past one decade, research in fake currency detection has shifted from classical image processing to shallow machine learning to deep learning architectures to hybrid artificial-intelligence systems and deployment-ready models for mobile and edge devices. This paper reviews recent progress in fake currency detection with emphasis on image-based methods using machine learning and deep learning. Further, challenges and future directions such as dataset standardisation, cross-currency robustness, adversarial resistance and explainability were discussed.