According to the Association of Certified Fraud Examiners, organizations worldwide suffer an estimated $3.7 trillion in losses annually due to fraud. Additionally, it is common for a typical company to lose five percent of its annual revenue as a result of fraudulent activities. As the prevalence of fraud continues to rise, it becomes increasingly crucial for industries to prioritize fraud detection. This course aims to demonstrate how analyzing historical data can help identify patterns of fraud and effectively combat it. To achieve this, the course introduces techniques derived from robust statistics and digit analysis. These methods enable the detection of unusual observations that are often indicative of fraudulent behavior. However, building a reliable tool for fraud detection poses two significant challenges. Firstly, the data used for training such a tool is often imbalanced or skewed, making it difficult to accurately identify fraudulent instances. Secondly, different types of misclassification carry varying costs, further complicating the development of a supervised fraud detection system. To address these challenges, the course presents a range of techniques designed to overcome data imbalance and cost disparities. Moreover, the course emphasizes the application of these techniques to both artificial and real datasets sourced from diverse fraud scenarios. By exploring a wide variety of fraud applications, participants gain a comprehensive understanding of how to effectively detect and combat fraud in their respective industries.