Health & Medicine, Australia (Commonwealth Union) – Scientists from the Macquarie University have shown a new way of connecting personal records and guarding privacy, a method with many possible applications across society. Its 1st utilization is for marking cases of rare genetic disorders.
In the United States, a five-year-old boy carries a rare genetic mutation in the GPX4 gene, a trait shared with a mere 10 other children globally. This genetic anomaly leads to abnormalities in both the skeletal structure and the central nervous system. Although it’s probable that several databases worldwide contain records of additional children with the same condition, the true extent of this disorder remains concealed due to legal and commercial privacy protections.
However, researchers pointed out a scenario in which information related to this condition could be gathered and quantified while still upholding privacy standards. A breakthrough in this direction has been achieved by researchers associated with the Macquarie University Cyber Security Hub. This team, led by Dr. Dinusha Vatsalan and Professor Dali Kaafar from the University’s School of Computing, collaborated with Mr. Sanath Kumar Ramesh, the father of the affected boy and CEO of the OpenTreatments Foundation based in Seattle, Washington.
The collaborative effort has resulted in the development of an innovative technique. This technique holds the potential to collect and aggregate data pertaining to the condition, all the while maintaining the confidentiality of individuals’ private information.
“I am very excited about this work,” said Mr Ramesh, whose foundation gave backing and commenced the project. “Knowing how many people have a condition underpins economic assumptions. If a condition was previously thought to have 15 patients and now we know, having pulled in data from diagnostic testing companies, that there are 100 patients, that increases market-size hugely.
“It would have a significant economic impact. The valuation of a company working on the condition would go up. Product costing would go down. How insurance companies account for medical costs would change. Diagnostic companies would target [the condition] more. And you can start to do epidemiology more precisely.”
The process of connecting and enumerating data records may appear straightforward; however, Professor Kaafar emphasizes that it encompasses numerous complexities. To begin with, the rarity of the disease means there is no centralized repository; instead, records are scattered across the globe. He further indicated that in this instance, they are dispersed across 100s of databases. Moreover, from a business standpoint, data holds considerable value, and the entities possessing it might not be inclined to share it willingly.
Beyond these challenges lies the technical hurdle of reconciling data that has been documented, encoded, and stored using disparate methods. This task also involves accounting for instances where individuals are counted more than once within and between various databases. And overarching all these obstacles are the critical concerns surrounding privacy. Professor Kaafar indicated that they are handling exceedingly sensitive health-related data.
In the past, this personal data was necessary not solely for a basic calculation of patient numbers and epidemiological insights, but also to guarantee the distinctiveness of records and their potential for linkage.
Dr. Vatsalan and her team harnessed a technique referred to as Bloom filter encoding combined with differential privacy. They devised a suite of algorithms intentionally designed to introduce a sufficient level of noise into the data. This noise obscures exact specifics to the extent that they cannot be extrapolated from individual records. Yet, it still enables the identification of patterns among records sharing the same disease condition, allowing them to be grouped together.
To validate the efficacy of their approach, they tested it using North Carolina voter registration data. The outcomes demonstrated that their method resulted in an almost negligible error rate, all the while ensuring an exceptionally high level of privacy, even when dealing with severely corrupted datasets. Importantly, their technique exhibits superior performance compared to existing methodologies.
