Enhancing Online Recruitment Fraud Detection Through Deep Learning And 2d Convolutional Neural Networks (CNN 2d)

Abstract

Online recruitment platforms have revolutionized the hiring process, but they have also given rise to fraudulent job postings, causing financial losses for job seekers. To address this issue, a deep learning-based methodology is proposed for detecting online recruitment fraud (ORF) using a novel dataset sourced from Fake Job Posting, Pakistan Job Posting, and US Job Posting datasets. The approach leverages Bidirectional Encoder Representations from Transformers (BERT) and Robustly Optimized BERT Pre-training Approach (RoBERTa) to transform job details into numerical vectors. To tackle the high class imbalance in the dataset, the SMOTE variant, SMOBD, is applied for effective class balancing. The experimental framework integrates these enhanced features with a two-dimensional Convolutional Neural Network (CNN2D) for job classification. Results demonstrate that the combination of BERT features and SMOBD with CNN2D achieves the highest classification accuracy of 98.68%. This methodology addresses the limitations of outdated datasets, providing a robust solution for detecting fraudulent job postings and significantly contributing to the prevention of online recruitment scams.