◇
Measure the trade-offs
Compare evaluated differential privacy, fully homomorphic encryption, and secure multi-party computation configurations, without treating their guarantees as equivalent.
Research · Software Engineering · Edge AI
Hi, I'm
Quoc Lap
Trieu.
Making intelligent systems more private, efficient, and practical at the edge.
PhD researcher at Western Sydney University, exploring secure machine learning inference across resource-constrained devices and distributed systems.
DATA → TRUST → INTELLIGENCE
SCROLL TO EXPLORESYDNEY, AUSTRALIA ↘
01 / ABOUT
Research with real-world
constraints.
Privacy guarantees are only meaningful when systems can deliver useful results within practical limits. My work examines how privacy, prediction quality, computation, communication, and energy interact at the edge.
03Research questions
SMCSecure multi-party inference
Edge AIResource-aware execution
02 / RESEARCH
Three connected
questions.
From understanding the privacy and resource trade-offs, to developing and evaluating complementary optimisations for secure inference.
⌘
Optimise the workload
Reduce secure-inference cost using accuracy-constrained numerical selection, preprocessing, authorised splitting, and adaptive scheduling.
✳
Optimise the protocol
Evaluate fused secure non-linear operations alongside straggler-aware party selection in distributed edge settings.
03 / FEATURED WORK
INTERACTIVE PHD PRESENTATION
Adaptive privacy-preserving
machine learning for
edge intelligence.
A visual walk-through of the problem, research questions, key experimental evidence, limitations, and future directions.
Open interactive slides ↗