Tegan Brennan
Stevens Institute of Technology, NJ·Computer Science·Early-career
Programs that handle private data can give secrets away without ever revealing the data directly. Attackers can instead watch side channels, observable byproducts of execution such as how long a computation takes.
- Fields
Computer Science
Engineering
- Topics
side-channel attacks
timing channels
just-in-time compilation
- Grants
- 2 known grants, through Sep 2030
What they work on
- Mark just-in-time compiler optimizations capable of introducing side-channel vulnerabilities, using taint analysis that tracks the flow of sensitive data through a program.
- Quantify the strength of compiler-induced side channels with an information-theoretic metric computed from profiling data the just-in-time compiler already collects, and use it to decide whether to apply or revert each optimization.
- Augment the compilation framework with machine-learning strategies that learn from historical data to predict the security impact of a given optimization before it is applied.
- Define a threat model for adaptive neural networks and build a machine-learning pipeline that exploits timing observations to recover information about private user inputs.
- Develop automated fuzzing to guide adaptive neural network development away from timing channels, plus an online monitoring framework checking privacy, accuracy, and performance in deployment.
More about the research
Two projects examine timing side channels in places where efforts to make software faster accidentally create them: inside compilers that translate code as it runs, and inside neural networks designed to use less memory and computation.
Why it matters
Both efforts aim to keep sensitive user data private without sacrificing the speed people expect. A privacy-aware compiler would relieve developers of guarding against leaks created downstream by optimization, while tools that detect and suppress timing channels could let efficient neural networks be deployed with more confidence on mobile phones and small smart devices that process personal data.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Isil Dillig
University of Texas at Austin, TX·Computer Science
Evaluate the security of JavaScript just-in-time (JIT) compilers through empirical study, to find which components are most open to attack and which matter most for browser speed.
Shiyi Wei
University of Texas at Dallas, TX·Computer Science·Early-career
Identify hidden relationships among the configuration options of Android taint analysis tools, and present them in a unified, configuration-aware interface that makes tuning easier.
Ruzica Piskac
Yale University, CT·Computer Science
Automatically locate power and electromagnetic side-channel vulnerabilities in low-level code by combining dynamic taint tracking and random testing with symbolic execution, satisfiability modulo theories solving and model counting.
The school
Stevens Institute of Technology: 4 labs on PI Pages, $71M a year on research, 72 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.