Scott Heidbrink

Alexis Cooper, Xin Zhou, Scott Heidbrink et al.

Software flaw detection using multimodal deep learning models has been demonstrated as a very competitive approach on benchmark problems. In this work, we demonstrate that even better performance can be achieved using neural architecture search (NAS) combined with multimodal learning models. We adapt a NAS framework aimed at investigating image classification to the problem of software flaw detection and demonstrate improved results on the Juliet Test Suite, a popular benchmarking data set for measuring performance of machine learning models in this problem domain.

Scott Heidbrink, Kathryn N. Rodhouse, Daniel M. Dunlavy

We explore the use of multiple deep learning models for detecting flaws in software programs. Current, standard approaches for flaw detection rely on a single representation of a software program (e.g., source code or a program binary). We illustrate that, by using techniques from multimodal deep learning, we can simultaneously leverage multiple representations of software programs to improve flaw detection over single representation analyses. Specifically, we adapt three deep learning models from the multimodal learning literature for use in flaw detection and demonstrate how these models outperform traditional deep learning models. We present results on detecting software flaws using the Juliet Test Suite and Linux Kernel.

Mark O'Neill, Scott Heidbrink, Jordan Whitehead et al.

We describe TrustBase, an architecture that provides certificate-based authentication as an operating system service. TrustBase enforces best practices for certificate validation for all applications and transparently enables existing applications to be strengthened against failures of the CA system. The TrustBase system allows simple deployment of authentication systems that harden the CA system. This enables system administrators, for example, to require certificate revocation checks on all TLS connections, or require STARTTLS for email servers that support it. TrustBase is the first system that is able to secure all TLS traffic, using an approach compatible with all operating systems. We design and evaluate a prototype implementation of TrustBase on Linux, evaluate its security, and demonstrate that it has negligible overhead and universal compatibility with applications. To demonstrate the utility of TrustBase, we have developed six authentication services that strengthen certificate validation for all applications.

Scott Ruoti, Jeff Andersen, Scott Heidbrink et al.

Secure email is increasingly being touted as usable by novice users, with a push for adoption based on recent concerns about government surveillance. To determine whether secure email is for grassroots adoption, we employ a laboratory user study that recruits pairs of novice to install and use several of the latest systems to exchange secure messages. We present quantitative and qualitative results from 25 pairs of novice users as they use Pwm, Tutanota, and Virtru. Participants report being more at ease with this type of study and better able to cope with mistakes since both participants are "on the same page". We find that users prefer integrated solutions over depot-based solutions, and that tutorials are important in helping first-time users. Hiding the details of how a secure email system provides security can lead to a lack of trust in the system. Participants expressed a desire to use secure email, but few wanted to use it regularly and most were unsure of when they might use it.