Affiliation:
1. Carleton University, Ottawa, Canada
Abstract
As one of the most pervasive current modes of communication, email needs to be fast and reliable. However, spammers and attackers use it as a primary channel to conduct illegal activities. Although many approaches have been developed and evaluated for spam detection, they do not provide sufficient accuracy. This deficiency results in significant economic losses for organizations. In this article, we first propose a framework for creating novel spam filters using Keras to combine a Convolutional Neural Network (CNN) with Long Short-Term Memory (LSTM) classification models. We then use this framework to introduce a specific solution applicable to realistic scenarios involving dynamic incoming email data in real-time. This solution takes the form of a real-time content-based spam classifier. We evaluate its performance concerning accuracy, precision, recall, false-positive, and false-negative rates. Our experimental results show that our approach can significantly outperform existing solutions for real-time spam detection.
Publisher
Association for Computing Machinery (ACM)
Subject
Computer Networks and Communications
Reference39 articles.
1. A review of the usage of machine learning in real-time systems;Abroyan N.;Proceedings of NPUA Information Technologies Electronics Radio Engineering,2016
2. Email classification using artificial neural network;Alghoul A.;Int. J. Acad. Eng. Res.,2018
3. A Survey of Phishing Email Filtering Techniques
4. Botnets threat analysis and detection;Atluri C.;Inf. Secur. Pract.,2017
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献