LEVERAGING DEEP LEARNING FOR DETECTING SUSPICIOUS ACTIVITY IN ONLINE FORUM DISCUSSIONS

Nowadays, people are passionate about using the internet in their daily lives, and this has resulted in the rapidly increasing adoption and use of online forums. An online forum can be defined as a medium to share one's thoughts, feelings and emotions towards specific multimedia artefacts such as pictures, videos and paintings etc. However, the use of online forums has leads to the execution of many illegal activities such as trading of black market money online, distributing copyrighted movies, as well as using illegal words. Therefore, Law enforcement needs a system to effectively deal with this problem. An example of such system include the Network Intrusion Detection System (NIDS) that assists system administrators detect network security vulnerabilities in their organization. On the other, many challenges arise when developing a flexible and effective NIDS for unplanned and unpredictable attacks. This paper proposes a deep learning-based approach to develop a flexible and efficient NIDS to analyze suspicious and criminal activities occurring in online forums. The proposed system combined a variety of Deep learning techniques and Natural Language Processing (NLP) for suspicious keyword extraction as well as Support Vector Machine (SVM) for detection and classification of suspicious keywords. We present the performance of our approach and compare it with some previous works. The metrics to be compared include accuracy, precision, recall, and f-measure values. This research improves system performance and security compared to existing systems