Secure AI Platforms: Building Resilient Machine Learning As a Service

Secure AI Platforms: Building Resilient Machine Learning As a Service

Programs Available

PhD 

Reference Number

20191855 

Opening Date

15 Aug 2019 12am AEST 

Closing Date

31 Oct 2019 12am AEDT 

Scholarship available to

Domestic 

Description

AI platforms and in particular ML as a service (predictive analytics) systems face several security and privacy challenges. First, AI platforms are increasingly deployed with publicly accessible query interfaces (Amazon, Google and Microsoft). These allow users to train models on potentially sensitive data and charge access to the models on a pay-per-query basis. If properly assessed, ML models may be deemed confidential due to the sensitive nature of their training data, the commercial value, or their use in security and defence applications. The ability to query these models makes the platforms vulnerable to information leakage where (i) the trained model can be ``stolen’’ and (ii) the sensitive data used for training the models can be recovered. Second, the nature of data modelling in AI includes inherent information stored as features (or layers) of the model. These features, when the ML model is exposed, can be exploited to breach individuals’ privacy. This project aims to  design privacy-preserving techniques for use as embedded within AI platforms where training data subjects are guaranteed privacy through on-device privacy-preserving AI and data owners are guaranteed that proprietary datasets and models remain private.

Availability

This scholarship is available to eligible domestic candidates to undertake a direct entry 3-year PhD program.

To be eligible for a scholarship, applicants are expected to have a record of excellent academic performance and additional relevant research experience and/or peer-reviewed research activity at a minimum level of a first-class Honours research degree or Master of Research, in line with the University’s scholarship rating guidelines. Refer to the Rating Scholarship Applicants section for more information about these guidelines.

Scholarship Components

The scholarship is comprised of the Tuition Fee Offset and a Macquarie University Research Excellence Scholarship (MQRES) Living Allowance/Stipend.

The stipend component of the scholarship is currently valued at AUD $27,596 per annum (2019 rate, tax exempt). The scholarship has been approved for a maximum of 3 years.

Contact Details

Name: Prof. Dali Kaafar and A/Prof. Mark Dras
Email: dali.kaafar@mq.edu.au, mark.Dras@mq.edu.au
Phone: 0435747249

How to Apply

For information on how to apply, select the "How to Apply" button below.

Please note, at the start of your application you should select the scholarship type of "HDR Project/Supervisor Specific Scholarship"

Back to the top of this page