E-MAIL CATEGORIZATION AND SECURE COMMUNICATION USING MACHINE LEARNING AND CRYPTO-STEGANOGRAPHY

Abstract

We are in an era of increasing cyber threats and information overload, the need for intelligent and secure digital communication systems has become paramount. Email, as one of the most widely used communication tools in both personal and organizational contexts, faces persistent challenges related to data privacy, message mismanagement, and unauthorized access. This study presents the development and evaluation of a secured, classified email system that integrates machine learning-based categorization with advanced crypto-steganographic techniques to enhance both the organization and security of electronic messages. The primary aim of the study is to design a system capable of automatically classifying emails according to their communication purpose - casual, official, or non-official emails and applying context-aware security measures to protect sensitive content during transmission. To achieve this, a synthetic dataset of 7,000 email was generated and structured using a comprehensive database schema that includes metadata such as sender, recipient, subject, body, and date. The Naïve Bayes algorithm was employed to classify emails based on textual features extracted through Term Frequency-Inverse Document Frequency (TF-IDF) vectorization.