Machine learning approach for threat detection on social media posts containing Arabic text

Recently, social media has become a part of daily people's routine. People frequently share images, text, and videos in social media (e.g., Twitter, Snapchat, Facebook, and Instagram). Consequently, there is a demand for an automated method to monitor and analyze the shared social media content. This research developed a method that aims to detect any threat in the images or comments in the shared content. Instagram has gained popularity as the most famous social media website and mobile application for media sharing. Instagram enables users to upload, view, share, and comment on a media post (image or video). There are many unwanted contents in Instagram posts, such as threats, which may cause problems for society and national security. The purpose of this research is to construct a model that can be utilized to classify Instagram content (images and Arabic comments) for threat detection. The model was built using Convolutional Neural Network, which is a deep learning algorithm. The dataset was collected utilizing the Instagram API and search engine and then labeled manually. The model used was retrained on the images and comments training set with the classes of threat and non-threat. The results show that the accuracy of the developed model is 96% for image classification and 99% for comment classification. The result of this research will be useful in tracking and monitoring social media posts for threat detection.