Privacy-Preserving Machine Learning Using Consortium Blockchain in Vehicular Social Networks

Abstract

It presents both opportunities for intelligent transportation systems and challenges for ensuring privacy during the data analysis process that VSNs (Vehicular Social Networks) generate. Our paper proposes a threshold Paillier cryptosystem and consortium blockchain for training Support Vector Machines (SVMs) on vertically partitioned datasets. Traditional approaches that rely on trusted third parties are insecure compared to blockchain-based collaboration. Most computations are performed locally and intermediate values are only shared if they are encrypted. This ensures high levels of privacy and efficiency. This model (PP-SVM) offers classification accuracy comparable to standard SVMs, resulting in privacy-preserving learning environments in virtual social networks. As a result of this approach, sensitive user data is effectively protected, and a robust sense of trust among network members is fostered. In addition to ensuring data integrity, consortium blockchain technology promotes collaborative learning by leveraging its inherently decentralized nature, facilitating secure interactions and shared decision-making processes. With the rapid evolution and increasing adoption of vehicular social networks, preserving user privacy has become increasingly crucial, demanding scalable and reliable security mechanisms.