Secure Systems Group

About Us

Secure Systems Group (SSG) is part CrySP at the University of Waterloo. SSG's focus is on understanding how to design and build systems that are simultaneously secure, easy to use, and inexpensive to deploy. This involves both building and evaluating systems but also occasionally breaking the security and privacy guarantees claimed by existing systems.

SSG came to Waterloo when Prof. N. Asokan moved from Aalto University to David R. Cheriton School of Computer Science in Fall 2019. You can see more information about our past projects at the Aalto SSG pages.

Our Research

Our current research interests fall into the following major themes: (for the time being, the links below take you to Aalto SSG pages).

• Platform security and applications: we investigate the design of new hardware and software platform security techniques as well as the use of current, widely deployed platform security techniques to secure applications and services.
• Machine learning and security/privacy: we study how to apply machine learning techniques to solve security and privacy problems as well as security and privacy challenges that arise in machine learning applications in general.

In addition to these major themes, we have several other themes where we do exploratory work.

Publications

Our publications from 2019 onwards can be found on the CrySP publications page. Earlier publications are available on the Aalto SSG publications page.

Sometimes we blog about our research results.

Dissemination

Below is a list of publicly available source code related to past and current SSG projects.

AD3	Adversarial input detector for deep reinforncement learning (DRL)
C-FLAT	Control flow attestation for embetded systems software
CONF-ML	Conflicts between ML protection mechanisms
DAWN	Dynamic adversarial watermarking of neural networks
HardScope	Run-time scope enforement on RISC-V
Intel SGX related projects	Projects related to Intel Software Guard Extensions (SGX)
Language data augmentation	Data augmentation techniques for toxic language classification
MiniONN	Privacy-preserving neural networks
Open-TEE	Virtual TEE compliant with the GlobalPlatform TEE specification
RecAgglo	Recursive Agglomerative Clustering (RecAgglo) for categorical data
ParChoice	Effective writing style transfer via combinatorial paraphrasing
Pointer Authentication	Projects related to ARMv8.3-PAuth
PRADA	Protection against DNN model stealing attacks
WAFFLE	Watermarking in federated learning