Research

The goal of PurS3 lab is to secure modern systems. To do so, we find, fix, and prevent vulnerabilities in both software and hardware. Consequently, we develop new systems that are secure by design. We are interested in all systems: Smartphones, IoT devices, Web Services, etc.

We strongly believe in diversity and aim to be a diverse group with members from different backgrounds.

Here are some themes and techniques that we currently work on:

Embedded Systems Security. We focus on various areas targeted towards improving the security of embedded systems, such as Vulnerability Detection and prevention, Developer tools to engineer secure systems.

Security of Continuous Integration Workflows. Continuous Integration (CI) has become essential to the modern software development cycle. Developers engineer CI scripts, commonly called workflows or pipelines, to automate most software maintenance tasks, such as testing and deployment. We focus on defining the desired security properties of a workflow and developing platform-independent techniques to verify and enforce the security properties.

Vulnerability Detection. We work on developing novel vulnerability detection techniques for software across the entire stack — applications, libraries, OS kernel, and drivers.

Retrofitting Memory Safety to C. Checked C extends legacy C with checked pointer types that are restricted by the compiler to safe uses. Checked pointer types can coexist with legacy pointers, thus admitting incremental conversion of legacy C to Checked C, in the style of migratory typing. We are working on a technique to port an existing C program to Checked C by converting its pointers to have checked type. More Details.

Machine Learning + Security. Machine Learning (and Deep Learning) provides us very useful techniques to achieve best-effort solutions in the presence of a large corpus of data. Many problems in security are proven to be intractable; however, providing a best-effort solution is reasonable. On this front, we are exploring the use of various Machine Learning techniques to handle the following problems:

  • Vulnerability Detection
  • Binary Analysis
  • Mobile Privacy

Fuzzing : New Frontiers. In general, fuzzing or automated dynamic testing of complex systems (IoT devices, Kernel Drivers, Network Programs, etc.) is still a very hard problem. We explore different techniques that will enable a security researcher to help the fuzzing techniques to test these complex programs effectively. We have exciting ideas in this direction. Get in touch to know more.

… and more.