Toby is a Professor whose research is focused on the problem of how to build highly secure computing systems cost-effectively. Prior to joining Melbourne in May 2016, he was employed in the Software Systems Research Group of NICTA (now Data61), and was a Conjoint Senior Lecturer in the school of Computer Science and Engineering of UNSW.
Security research at the University of Melbourne.
We work across cryptography, formal methods, software and systems security, privacy, and trustworthy machine learning.
People
Teaching and research faculty
News
Recent updates from the group
Xingliang Yuan and collaborators revisit DP-SGD at ACM CCS
Wenhao Wang, Shujie Cui, Hui Cui and Xingliang Yuan revisit the assumptions behind privacy guarantees for differentially private stochastic gradient descent. Their paper appears in the ACM CCS 2026 programme.
Following Dragons accepted to ICSE 2027
Viet Hoang Luu, Amirmohammad Pasdar, Wachiraphan Charoenwet, Toby Murray, Shaanan Cohney and Van-Thuan Pham use insights from code review to direct automated software testing. The paper will appear at ICSE 2027.
Aaron Bembenek begins a McKenzie Postdoctoral Fellowship
Aaron began a three-year University of Melbourne fellowship in July to develop AI systems that combine formal reasoning with the strengths of learned models.
Rethinking how we measure adversarial risk in generative AI
Andrew Cullen, Neil Marchant, Paul Montague, Jiani Xie and Benjamin Rubinstein argue for evaluating AI security against realistic misuse scenarios and system-level risks.
Benchmarking tools that check constant-time code
Zhenzhi Lai, Zhiyuan Zhang, Tian Qiu, Van-Thuan Pham and Udaya Parampalli benchmark seven tools for checking whether cryptographic code follows the constant-time programming discipline.
Writing
Public writing and commentary
A critical flaw affecting 70% of all websites has the cybersecurity world panicked
What a PHP security flaw and the rush to patch it reveal about the maintenance of widely used web infrastructure.
How Quantum Tech Could Revolutionise Military Logistics and Reconnaissance
The authors explain where quantum optimisation might help with logistics and what limits its use today.
Big AI wants to slow down AI research. Is it a safety pause or a strategic retreat?
Andrew examines the safety and commercial incentives behind calls for an AI research pause and the challenge of making it stick.
Technology for Real Impact: Scaling the Digital Economy through IOWN and Data & AI Infrastructure
Benjamin reflects on the panel’s discussion of responsible AI, digital infrastructure and collaboration between universities and industry.
Don’t Use Ordinary Software to Contain Software-Hacking Agents
Toby explains why evaluating AI agents that can find software flaws calls for stronger isolation and more careful evaluation design.
Australian musicians hate AI using their songs, but have little legal protection
Andrew looks at AI training, copyright and what transparency and fair remuneration could mean for Australian musicians.
People