Toward Keyword Extraction in Constrained Information Retrieval in Vehicle Social Network

• Published in: IEEE transactions on vehicular technology Vol. 68; no. 5; pp. 4285 - 4294
• Main Authors: Yang, Zhen, Yu, Haiyang, Tang, Jiliang, Liu, Huan
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.05.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Cloud computing; Constrained information retrieval; content analysis; Data processing; Encryption; In car entertainment; Indexes; Information management; Information retrieval; keyword extraction; Keywords; Retrieval performance measures; Social networking (online); Social networks; Spatial distribution; Vehicle safety; vehicle social network
• ISSN: 0018-9545; 1939-9359
• DOI: 10.1109/TVT.2019.2906799

With the emergence of various vehicular apps for automotive navigation, driving safety, and in-car entertainment system, vehicular social data is starting to explode. Such an explosion has posed a great challenge for a centralized pattern of computing, storing, and managing information, calling for a more decentralized (cloud-based) style of processing data for vehicular social network. Decentralized information processing requires encryption of sensitive information prior to out-sourcing for the purpose of protecting data from unsolicited access and thus a searchable keyword index needs to be extracted from the encrypted document for solicited access. However, most existing searchable encryption schemes, based on the words in the subject or content, suffers from poor retrieval performance. In this study, a novel keyword extraction metric based on spatial distribution of a particular text is proposed with the view to improving the retrieval performance. In the proposed solution, constrained information retrieval in vehicle social network is conceptualized as retrieval of data with extremely strict security and privacy preserving policies and a cloud-based vehicle social information retrieval framework is developed. In line with the framework, experiments have been conducted on benchmark data sets and diverse evaluation metrics. The results suggest that the solution proposed in this paper is capable of a wider application and a better performance.