Detection of Spammers Using Modified Diffusion Convolution Neural Network
Autor: | Wenxin Liang, Hui Li, Zihan Liao |
---|---|
Rok vydání: | 2020 |
Předmět: |
Structure (mathematical logic)
021110 strategic defence & security studies Diffusion (acoustics) Social network business.industry Computer science 0211 other engineering and technologies 02 engineering and technology Information security Machine learning computer.software_genre Convolutional neural network Regularization (mathematics) Spamming 0202 electrical engineering electronic engineering information engineering Graph (abstract data type) 020201 artificial intelligence & image processing Artificial intelligence business computer |
Zdroj: | Lecture Notes in Computer Science ISBN: 9783030604691 |
Popis: | Social network brings convenience to our life, but it also provides a platform for spammers to spread malicious information and links. Most of the existing methods for identifying spammers mainly rely on the user’s behavior information to learn classification models. But, privacy and information security issues make it impossible to monitor all behavioral of users. In addition, owing to the diversity and variability of spammers’ strategies, it is difficult to distinguish them from legitimate users only by their own behavior. To solve this challenge, in this paper we propose a novel spammer detecting method using DCNN (Diffusion Convolution Neural Network) which is a graph-based model. And DCNN model can learn behavior information from other users through the graph structure (i.e., social network relationships). However, the original DCNN model is a general classification model. In order to make the original DCNN model more effective for detecting spammers, we modify it by using attenuation Coefficient and social Regularization, which is called DCNN+CR model. The experimental results on real-world Twitter dataset show that the proposed DCNN+CR model outperforms existing methods, especially in terms of accuracy and F1-score. |
Databáze: | OpenAIRE |
Externí odkaz: |