Network Spam Detection Based on CNN Incorporated with Attention Model

Abstract

The rapid popularization of computer technology and Internet communication has not only brought convenience to people's life and work, but also brought many new network security challenges, such as malware, Trojan horse and spam. Among them, network spam is the preferred attack medium for network criminals to launch malicious activities. It usually includes phishing links, malicious warnings, and viruses. Therefore, fast and efficient spam detection technology has gradually become a research hotspot of network security. However, at present, the sending speed and scale of online mail are growing, the traditional network spam detection methods cannot meet the needs of users. With the in-depth development of machine learning, intelligent spam detection technology has been continuously applied, but the traditional machine learning methods often rely on the extraction of various features, which is time-consuming and difficult. To solve the problem, this paper, by taking advantage of the benefit of deep learning that can be completed automatically in feature extraction, proposes a CNN incorporated with attention model for network spam detection, including network spam collection, data preprocessing by using Glove model to train word vector, and model training. The experiments have verified the effectiveness of the proposed method.