Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{206315,
        author = {Mohammed Faiz and A. D. Wakhare},
        title = {Detecting Fraudulent Job Postings Using Bidirectional LSTM and SMOTE-Based Class Balancing},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {2},
        pages = {914-925},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=206315},
        abstract = {The widespread proliferation of online job platforms has brought with it an alarming rise in fraudulent job advertisements that exploit job seekers through identity theft, financial scams, and phishing attacks. Traditional rule-based detection mechanisms are no longer capable of keeping pace with the sophistication and volume of modern recruitment fraud. This paper presents Recruit Guard, an intelligent deep learning-based framework for automated detection of fraudulent job postings using Natural Language Processing (NLP) and Bidirectional Long Short-Term Memory (Bi-LSTM) networks. The proposed system ingests job posting data comprising textual fields such as job description, company profile, employment requirements, benefits, and salary information, and processes them through a structured NLP pipeline involving text cleaning, stop-word removal, tokenization, lemmatization, and sequence padding. To address the inherent class imbalance in the dataset where genuine postings far outnumber fraudulent ones—the Synthetic Minority Oversampling Technique (SMOTE) is applied prior to model training. The core classification model employs a Bi-LSTM architecture that processes token sequences in both forward and backward directions, enabling richer contextual understanding of job descriptions than unidirectional LSTM models. Extensive experimentation on the benchmark Fake Job Postings Dataset demonstrates that the proposed Bi-LSTM model achieves a classification accuracy of 97.2%, with precision of 96.8%, recall of 97.5%, and an F1-score of 97.1%. These results outperform traditional machine learning baselines including Logistic Regression, Naive Bayes, Decision Tree, Random Forest, and standard LSTM. Additionally, the system is deployed as a web-based application that allows users to verify job authenticity in real time by entering relevant details. This work contributes a practical, scalable, and effective solution for protecting job seekers in the increasingly risky landscape of online recruitment.},
        keywords = {Fake Job Detection, Recruitment Fraud, Bidirectional LSTM, Natural Language Processing, SMOTE, Deep Learning, Online Job Portals, Binary Classification, Web-Based Prediction, Text Classification.},
        month = {July},
        }