Article Preview
TopIntroduction
With the rapid development of information technology, the Internet has long penetrated all aspects of people’s lives. According to Hootsuite data, as of January 27, 2022, the number of social media users is 4.62 billion, an increase of 424 million (10%) over the same period last year (Social, 2022). The development of the Internet has brought people convenience, but also led to some problems. The emergence of increasingly unhealthy webpages harms social stability and people’s physical and mental health. The discovery of unhealthy webpages is an important way to solve the stable development of society, and this has been the focus of a growing research stream. Oram et al. (2021) developed a phishing web detection model based on a light gradient booster that uses phishing websites to monitor and filter phishing webpages by mimicking URLs. Namara et al. (2018) investigated how to adapt Facebook’s privacy features to users’ personal preferences based on user-defined privacy. Ertam (2018) used web scraping to collect classified news headlines and summaries from a news agency website and classified the test data using vector learning and deep learning methods. Yizhi Liu et al. (2021) developed an efficient mobile malicious webpage detection framework based on deep learning and edge clouds, which applies the ideas of edge computing and multi-device load optimization to MMWD, which can optimally deploy multiple device resources and detect mobile malicious web pages more effectively. Sajedi (2019) used an integrated algorithm to assign weights to weak classifiers. Then, he used a genetic algorithm to select the optimal weak classifier member and set optimal settings for optimal integration. Patil and Patil (2019) proposed using feature selection methods and machine learning to detect malicious web pages to detect unhealthy webpages.
Although the current research has achieved numerous results, the existing results are mainly the application of traditional machine learning or pattern matching methods, and there are problems such as a high false positive rate, high false negative rate, and high labor cost. Therefore, this paper constructs an unhealthy web page discovery system based on a convolutional neural network. A web crawler crawls the key text information in unhealthy web pages, and the key information crawled is filtered and screened by a convolutional neural network, which greatly improves the recognition rate of unhealthy web pages, helps to purify cyberspace, and improves the network environment.
In addition, we use some abbreviations, such as DQN (Deep Q-Network), PPO (Proximal Policy Optimization), URL (Uniform Resource Locator), CNN (Convolutional Neural Network), SMTP (Simple Mail Transfer Protocol), MD5 (Message-Dest algorithm), and SVM (Support Vector Machines).