| Abstract |
Since the creation of the cloud computing (CC) paradigm, there has been a rising interest in the adoption of CC from municipalities and city governments toward their effort to address complex civilian issues. CC is a prolific, service-based delivery of enormous computer processing power and data storage across connected communications channels, and is the most significant technological advancement in the IT industry since the inception of the Internet (web), and more so in the evolution of smart cities. It became feasible to share confidential data records through CC for further analysis and mining. However, data sharing prompted various security threats and concerns that have been imposed via ever-escalating phishing attacks using advanced deception. Supreme cyberspace security solutions and mitigation against such phishing attacks became mandatory due to the momentous impacts on global security and the economy. Intensifying the protection performance, precision, and effectiveness of data mining against such cyber-forgery by lowering computational costs appears challenging. Thus, privacy-preserving data mining has emerged as a new research avenue, where various algorithms are developed to anonymize the data to be mined. The K-anonymizing privacy-preserving approach, being the most prospective one, is widely used to secure data. It preserves published data from being linked back to an individual. Yet, the protection and truthfulness potency of this generalized technique is limited to a tiny output space, and often leads to unacceptable utility loss in cases of strict privacy requirements. To surmount this limitation, we propose a hybrid K-anonymity data relocation algorithm. The data relocation, which is a tradeoff between trustworthiness and utility, acts as a control input parameter. The performance of each K-anonymity’s iteration is measured to decide the feasibility of data relocation, where data rows are changed into small groups of tuples. These tuples, being indistinguishable to each other, create anonymizations of finer granularity with an assured privacy standard. Furthermore, this approach is designed for L-diversity, and (a, K) anonymity. Experimental results demonstrate considerable utility enhancement as a function of a relatively small number of group relocations. © 2019 Elsevier Inc. All rights reserved. |
