WResM-Net: Blockchain and WideResnet Maxout Network for Attribute Matching in Social Networks

Abstract

Nowadays, social networks are increasingly popular because of the excessive growth of online interactions and the increasing advancement of mobile applications. The attributes of the user are generally used in social networks for establishing social relationships with other people with similar attributes. Meanwhile, social networks encounter privacy leakage issues because of illegal disclosure of users' private data and threaten data confidentiality. Attribute matching is performed to compare the similarity among users to handle the leakage of user information. The prevailing techniques employed for user attribute matching don't validate the user behavior information from multiple virtual accounts. Thus, a novel deep learning model WideResMaxout-Network (WResM-Net) is introduced in this research to perform attribute matching in blockchain-based social networks. The attribute matching is performed under two phases, namely system initialization and user communication, where the matching task is executed by considering different entities. Here, the attributes are normalized using the log scaling method, then the relevant attributes are selected using Canberra distance, and attribute matching is finally performed using WResM-Net. In addition, the superiority of WResM-Net is validated by comparing it with existing matching schemes, where the WResM-Net attained a matching score, computational time, and memory of 91.876%, 25.987 ms, and 19.776 Mb.