Unsupervised Anomaly Detection Using Machine Learning

Abstract

In current era, Internet has become the essential component of life. The Internet is one of the most influential innovations in recent history. Though most people use the Internet for productive purposes, some use it as malicious intent. The Internet and the computers connected to it increasingly become more enticing targets of attacks, As the Internet links more users together and computers are more prevalent in our daily lives. Computer security often focuses on preventing attacks using usually authentication, filtering, and encryption techniques, but another important facet is detecting attacks once the preventive measures are breached. Prevention and detection complement each other to provide a more secure environment.

Fraud detection methods are continuously developed to defend criminals in adapting to their strategies. Fraud detection techniques quickly identify frauds. Here, for credit card frauds, clustering approach is used. Data is generated randomly and then for detecting the transaction, K-means clustering algorithm is used. Clusters are formed to detect fraud in transaction which are low, high, risky and high risky. K-means algorithm is simple and efficient for credit card fraud detection. Clustering and related techniques have been used to locate anamolies in a dataset. The algorithms are implemented in Python language using the Spyder software provided by the Anaconda Distribution Platform. The results show that Isolation Forest algorithm for unsupervised anomaly detection have better accuracy compared to K-Means algorithm. As Credit card has the power to purchase the things, its frauds also increased.