Spam Detection in Chat Application

Abstract

Emails, Chats, SMS are widely used in this day and age for communication. But this rise in technology has invited spammers. Text spams are used in SMS, chats, emails, etc. Spam detection in messages has been since a long time. However nowadays these spams are increasing in the form of images. This project displays an application of Naive Bayes for text-based Spam classification and Deep CNN for image-based Spam classification in our own chat application. In addition to these, another method of Image spam classification is used, using Tesseract for Object Character Recognition technique and feeding the text extracted to our text based classifier. This proposed approach in text based spam classification has achieved 97% using Naive Bayes classifier. The CNN model achieved 96.87% accuracy for image spam classification. The app takes less than a second to send text with spam classification feature and took around five seconds to send image with spam classification in web. However in mobile, it takes slightly more time for image spam classification (1 minute and 10 seconds).