There is a growing interest from the adopters of Internet of Things (IoT) to realize business insights and tangible advantages associated with analyzing IoT data using Machine Learning (ML). Considering that IoT devices have constraints on their resources, data used to train ML models is often stored and processed on a cloud service provider infrastructure. While moving these critical workloads to cloud renders a business advantage, it also presents an opportunity for data breaches and threats to data privacy. Hence, it is a good practice to perform secure computation on the raw data generated at the IoT (and/or edge) devices into privacy preserving data before further securely processing this data by the ML algorithms in other parties for insights to drive other applications and services.

The aim of this project is to investigate for an efficient secure computation method, which can preserve the privacy of the data generated by the IoT devices so that it is independent of the ML model utilized by the different parties involved to process the data. This thesis plans to investigate for a solution based on the latest development of secure computation methods for a right balance between data privacy and resource efficiency for the ML algorithms. The research outcomes consist of a proven privacy preserving method, a prototype system or framework which implements this method on IoT and/or edge devices, a demonstration of the application for popular ML algorithms and a comparative study with other generic privacy preserving ML algorithms when adapted for IoT applications.

The scholarship is funded through the Cyber Security Cooperative Research Centre (CSCRC), which is focused on delivering industry-driven cybersecurity research outcomes that have an impact and address real-world cybersecurity problems with innovative solutions. The CSCRC has been granted $50m of funding over 7 years from the Australian Government's Cooperative Research Centres Program.

The student will be supervised by a capable team including academics from UNSW and researchers from CSIRO's Data61. There will be an opportunity to engage with industry partners.

The scholarship is open to both domestic and international students. International applicants should be able to secure a Tuition Fee Scholarship from UNSW. This entails an academic record that is equivalent to an Australian First Class Honours degree in Computer Science from a reputed institute. A publication track record in machine learning and/or security is highly desirable.

The scholarship includes:

$37K per annum (tax-free) for a period of 3 years
Access to operational budget for research support and conference travel
Mentoring and training opportunities through the CSCRC
Interested candidates should send their CV (less than 2 pages) and academic transcripts to Associate Professor Wen Hu (wen.hu@unsw.edu.au). 返回小木虫查看更多