|
[Print-friendly version]
Zim Chan is undertaking research into the use of agent technologies to enhance
the provision of safe, secure and depersonalized health data. This will enhance
CSIRO's privacy preserving Health Data Integration software tools. This project,
supervised by Prof Peter Croll [QUT], Dr Anthony Maeder [CSIRO] and Dr David
Hansen [CSIRO] is funded by CSIRO/Queensland Health’s E-Health Research
Centre (www.e-hrc.net).
Privacy concerns continue to be a major issue in any undertaking involving
health data, in particular patient records. Traditionally, techniques such as
de-identification and aggregation have been used to sanitize sensitive information.
There has also been considerable work in the area of privacy-preserving data
mining and secure multiparty computation in allowing for results to be generated
while mitigating information release. However, existing techniques are extremely
task-specific, and as health data is being explored in increasingly complex
ways, these techniques may not be appropriate, and de-identification alone is
no longer sufficient. Patient information can be subject to (often trivial)
re-identification attacks based on linkage with other datasets or other background
knowledge. Aggregation can also have a detrimental impact on the quality of
data.
The purpose of this research is to investigate the suitability of using intelligent
agents to mine distributed health data in a privacy-preserving manner. By allowing
agents to enter protected databases and act on behalf of users, users will be
able to retain some control over how the data is manipulated, despite not having
access to the actual data itself. This is significant because it will allow
for the future exploration of data in currently unanticipated ways, while still
protecting privacy constraints. At the same time, agents - distributed over
multiple databases - cooperate with one another to ensure that only information
which does not breach privacy constraints is released. This could be achieved
by the 'sanitization' of datasets via k-anonymization or micro-aggregation techniques,
or results generated directly from privacy-preserving data mining. A multi-agent
systems approach will study the distributed coordination, negotiation, planning
and execution methods to be used towards this end.
Related Publications:
Zim Chan, PR Croll, David Hansen and Anthony Maeder, The Use of Agent Technologies
for Preserving Privacy with Health Records when Linking Federated Databases,
accepted to appear at the Health Informatics Conference Sydney, Aug. HIC 2006.
|
