Using Contextual Features for Online Recruitment Fraud Detection

The recent growth of online recruitment and candidate management systems has established yet another media for fraudsters on the internet. The ever-growing size of the candidate pool has forced different industries to move to web-based candidate management systems. The advantages of such web-based systems are substantial. On one hand, they are the best means to filter through thousands of applicants for employers and on the other hand, the candidates find themselves in a convenient position while applying for a position. People with fraudulent motivations explore these systems to lure candidates in a hoax and extract sensitive information (e.g. contact information) using fake job advertisements. In this paper, we analyzed a publicly available dataset and used machine learning algorithms to classify job postings as fraudulent or legitimate. The contribution of this research is the inclusion of contextual features in the feature space, which revealed compelling improvements of accuracy, precision and recall