Fake Currency Detection Using Machine Learning and Image Processing

Counterfeit currency is a serious financial crime that can shake a whole country's economy; it can be prevented, but people are often unaware of how to tell a counterfeit note from a genuine one, so spreading awareness is key. This paper presents a Fake Currency Detection system that applies machine learning and image processing to classify banknote images as genuine or fake. Statistical features (variance, skewness, kurtosis, and entropy) are extracted from the images using wavelet transforms, and classification is performed by supervised models—Random Forest, K-Nearest Neighbours (KNN), and Gradient Boosting—combined in an ensemble (Voting Classifier) and tuned with GridSearchCV and cross validation. The paper reports a twenty-study literature survey, the software requirements specification (SRS), a layered, modular system design, the Waterfall development methodology, the implementation stack (Python, OpenCV, NumPy, Pandas, Scikit-learn, PyWavelets, Flask/Django, MySQL/MongoDB), and the testing and deployment strategy. Convolutional neural networks (CNNs), which excel at image pattern recognition, are discussed as a complementary direction. The source work also acknowledges limitations: model effectiveness depends on the quality and omprehensiveness of the training data, counterfeiters constantly adopt new techniques so the dataset must be kept up to date, and the models themselves must be secured against manipulation. Future directions include transfer learning, which adapts pre-trained models to reduce training time and resources, and generative adversarial networks (GANs) for building more comprehensive training datasets.