Fraud Detection in Current Scenario, Sophistications and Directions: A Comprehensive Survey

Abstract

Fraud Detection is one of the oldest areas of research. The requirement of an effective system that detects frauds effectively with zero loss exists until now. This is due to the increase in the technology, that influences both the ends; the user and the fraudster. Hence it becomes mandatory that the users need to stay a step ahead in this scenario. This paper discusses the changes that had taken place in the area of fraud detection. The flow of research from data mining approaches to machine learning approaches that were developed to defy the attacks are discussed here. It discusses the evolution of heuristic based mechanisms and graph based technologies that emerged in recent years. Further, it also discusses the need for Big Data based analysis in this domain. A few case studies are also discussed here to enable better understanding. Research challenges that exist in this domain in the current scenario are discussed along with the research directions.

A Review on Computer Technology Applications in Fraud Detection and Prevention

“…According to Kavitha & Suriakala, (2015), the Negative Selection algorithm was parallelized in the Hadoop environment to determine accuracy. It used the basic format of determining the outliers using the Euclidean distance and compared it with the threshold.…”

“…The original fraud detection systems which are now in existence use data mining and statistics to extract fraudulent information from available data. When fraud becomes more complex, these methods have become outdated which leads to the emergence of advanced detection mechanisms to improve its performance (Kavitha & Suriakala, 2015). There are plenty of tools used for fraud detection, each of which tries to retain maximum quality of service while keeping false alarm at a minimum rate.…”