Now showing 1 - 2 of 2
  • Publication
    Protecting organizational data confidentiality in the cloud using a high-performance anonymization engine
    Data security remains a top concern for the adoption of cloud-based delivery models, especially in the case of the Software as a Service (SaaS). This concern is primarily caused due to the lack of transparency on how customer data is managed. Clients depend on the security measures implemented by the service providers to keep their information protected. However, not many practical solutions exist to protect data from malicious insiders working for the cloud providers, a factor that represents a high potential for data breaches. This paper presents the High-Performance Anonymization Engine (HPAE), an approach to allow companies to protect their sensitive information from SaaS providers in a public cloud. This approach uses data anonymization to prevent the exposure of sensitive data in its original form, thus reducing the risk for misuses of customer information. This work involved the implementation of a prototype and an experimental validation phase, which assessed the performance of the HPAE in the context of a cloud-based log management service. The results showed that the architecture of the HPAE is a practical solution and can efficiently handle large volumes of data.
  • Publication
    A Systematic Comparison and Evaluation of k-Anonymization Algorithms for Practitioners
    The vast amount of data being collected about individuals has brought new challenges in protecting their privacy when this data is disseminated. As a result, Privacy-Preserving Data Publishing has become an active research area, in which multiple anonymization algorithms have been proposed. However, given the large number of algorithms available and limited information regarding their performance, it is difficult to identify and select the most appropriate algorithm given a particular publishing scenario, especially for practitioners. In this paper, we perform a systematic comparison of three well-known k-anonymization algorithms to measure their efficiency (in terms of resources usage) and their effectiveness (in terms of data utility). We extend the scope of their original evaluation by employing a more comprehensive set of scenarios: different parameters, metrics and datasets. Using publicly available implementations of those algorithms, we conduct a series of experiments and a comprehensive analysis to identify the factors that influence their performance, in order to guide practitioners in the selection of an algorithm. We demonstrate through experimental evaluation, the conditions in which one algorithm outperforms the others for a particular metric, depending on the input dataset and privacy requirements. Our findings motivate the necessity of creating methodologies that provide recommendations about the best algorithm given a particular publishing scenario.