Privacy-preserving healthcare informatics: A review

Posted on April 12, 2021 By LabLynx Journal articles

Last week's article looked at inference attacks in the cloud, but they happen in other ways as well. In this 2021 paper, Chong reviews the available literature on inference and other privacy attacks that occur on published healthcare information originally sourced from electronic health records (EHRs) and other health informatics systems. Ethical sharing of this kind of data with researchers conducting statistical analysis, improving clinical decision making, etc. is important, but it must preserve the privacy of the underlying individuals. After briefly discussing data publishing with a strong privacy focus, Chong looks at how healthcare data is typically stored and what threats exist against such data. The author then examines two well-established privacy models, including their strengths and limitations, that can be used to limit those threats: data anonymization and differential privacy. Despite these models, Chong closes by noting that "preserving privacy in healthcare data still poses several unsolved privacy and utility challenges" and expounds on areas for future research into those challenges.

Secure data outsourcing in presence of the inference problem: Issues and directions

Posted on April 7, 2021 By LabLynx Journal articles

Inference attacks occur when a (typically) malicious actor is able to infer from public information more sensitive information about a database or individual without directly accessing the original database or sensitive user information. That these sorts of attacks can be successful against applications and data on cloud-based infrastructure should not be a surprise to seasoned IT veterans. Recognizing these concerns, Jebali et al. examine the state of inference attacks in the cloud, the research that has been conducted so far on such attacks, and what can be done about them. Aside from examining a variety of security concerns in both communicating and non-communicating servers, the authors also address the inference problem, offering a potential solution that will in theory "optimize data distribution without the need to query the workload, then partition the database in the cloud by taking into consideration access control policies and data utility, before finally running a query evaluation model on a big data framework to securely process distributed queries while retaining access control." They conclude by recognizing more research must be done to test their proposed solution, and noting that past research has often overlooked other types of inference sources such as inclusion dependencies, join dependencies, and multivalued dependencies, which should be further examined.