Online Recruitment Fraud Detection (ORF) Using Deep Learning Approaches

Online Recruitment Fraud Detection (ORF) Using Deep Learning Approaches

K. Sravanthi¹, Y. Bhavani², A. Nikitha³, Ch. Murali Krishna4

1,2,3 UG Scholars, 4Assistant Professor

1,2,3,4 Department of CSE [Artificial Intelligence & Machine Learning],

1,2,3,4 Guru Nanak Institutions Technical Campus, Hyderabad, Telangana, India

Abstract – In this work, a new and effective deep learning method for identifying online recruitment fraud (ORF) through advanced natural language processing and transformer-based models. Unlike conventional approaches that rely on outdated datasets or single-model pipelines, our method integrates a comprehensive dataset aggregated from multiple platforms and addresses the challenge of class imbalance using ten sophisticated variants of the Synthetic Minority Oversampling Technique (SMOTE). The architecture employs a dual-model strategy, evaluating both BERT and RoBERTa for high-precision classification, while also leveraging LSTM networks for sequential learning of job description patterns. This combination ensures improved detection accuracy without significant computational overhead, rendering it appropriate for use in real-time scenarios. Extensive experiments demonstrate the superiority of our SMOBD-SMOTE-enhanced BERT model, achieving up to 97% accuracy and recollection. Results confirm potential our unified approach in establishing scalable, adaptable, and high-performance solutions to combat online recruitment fraud effectively.

Key Words: Employment scam, Machine learning, Deep learning, Online recruitment, Fraud detection, Class imbalance, Data augmentation, SMOTE, Transformer-based models