A Social Communication Platform with Automated Toxicity Detection, Image Moderation and Secure Account Verification

Online social platforms still depend largely on user reports to find abusive messages and explicit images, so harm is often seen before it is removed. This paper presents a secure social communication and content-sharing platform that moderates content automatically at the time it is sent. Text and voice messages are screened by an NLP-based toxicity classifier (voice is first converted to text), and the same screening is applied to post comments. An escalating policy issues warnings after the first and second violations and blocks the account after the third. Uploaded images pass through an ML-based nudity detector, and an MTCNN-based face-recognition module finds registered users in uploaded photos and asks each of them to approve or reject the post. Registration is protected with email OTP verification and a duplicate-account check based on the profile picture. The platform is built with Django, MySQL and a Flutter/web front end. We describe the architecture, the moderation policy and the evaluation protocol, and we discuss the privacy and abuse risks of the face-based features.