Principal Investigator on DO NO HARM (a system to remediate vulnerabilities in medical devices), and subcontractor for HRL's MINDSET (a system to identify vulnerabilities in medical devices).
Extended VulChecker to identify a wide variety of CWEs in firmware binaries.
Helped write this paper and performed some analyses.
Publication: Identifying Behavior Dispatchers for Malware Analysis (AsiaCCS 2021)
DeepReflect - 2021
Trained a deep learning model to detect malicious functions within malware binaries using instruction- and CFG-based features. It is particularly useful when symbols and strings are missing, forcing the analyst to otherwise execute the malware in a dynamic sandbox.
Publication: DeepReflect: Discovering Malicious Functionality through Binary Reconstruction (USENIX 2021)
Techniques: Python, BinaryNinja, Scikit-learn, Keras
D3 (TII drone research) - 2021-2022
Assisted a post-doc and PhD student in designing a drone intrusion detection methodology based on external environmental data (e.g., sound of the propellers spinning).
Soldered and installed a Pi-connect, Raspberry pi, and microphone array onto the drone to collect telemetry data.
Assisted in the design and development of MLSploit, a flexible framework enabling machine learning model training and generating attacks to evade those models.
Utilized Python, Scikit-learn, and Keras to create various malware detection models and employed binary rewriting for executable malware that dynamically evades detection.
Collaborated with a research group to develop and test a novel method for detecting malicious intrusions into computers, involving Linux kernel modification and rootkit development.
Publication: Beholder: Phase-Space Detection of Cyber Events (2013)
Techniques: C
Oak Ridge National Laboratory - 05/2012-08/2012
Developed a JavaScript-based API for flexible scatterplot creation, to be utilized in a Human-Computer Interaction (HCI) study for Centers for Medicare & Medicaid Services (CMS).
Designed and developed visual interfaces for projects analyzing computer network traffic for malicious and anomalous patterns.
Publications:
situ: Situational Understanding and Discovery for Cyber Attacks (2012)
NV: Nessus Vulnerability Visualization for the Web (VizSec 2012)
Techniques: JavaScript, CSS, HTML
Oak Ridge National Laboratory - 05/2011-07/2011
Collaborated with the U.S. Marines and Centers for Medicare & Medicaid Services (CMS) to develop maintenance durability and cost/progress visualizations using Protovis, a JavaScript visualization library.