Multimodal Spam Classification Using Deep Learning Techniques

The internet has been beneficial to the society in ways more than one, the power to learn anything anywhere, the power to always be connected to the people you love. But as usual, there are two sides to coin. The E-mail system has been the backbone for communication between professionals for a very long time, but it is plagued by the unwanted influence of spam. In this paper we classify a mail into spam or not-spam (ham) by analyzing the whole content i.e. Image and Text, processing it through independent classifiers using Convolutional Neural Networks. We finally propose two hybrid multi-modal architectures by forging the image and text classifiers. Our experimental results outperform the current state-of-the-art methods and provide a new baseline for future research in the field.