Research on Malicious Words Detection based on Neural Network

At present, malicious words spreading not only damages the network environment but also bring an unsatisfied experience for all Internet users. Therefore, immediately detecting the malicious words becomes an important issue for existing Internet social networks and can assist the administrator to dispose the emergency issues. Traditional methods are primary concentrated on the several certain key words extraction, which may identify wrong malicious words and cost numerous computation costs. Subsequently, machine learning is utilized to detect malicious words, which enhances the identification accuracy with trained machine learning model. In this paper, we establish a neural network model to achieve malicious words identification with reasonable computation costs. We initially extract the sentence features through a multiple-layer perceptron and divide the malicious features to a trained neural network. Indeed, the features are divided to precisely identify the malicious words. From our extensive experimental results, we can conclude that our proposed methods can automatically identify the malicious words with more than 85% detection accuracy.