icon-carat-right menu search cmu-wordmark

Research Review 2022

At the 2022 Research Review, our researchers detail how they are forging a new path for software engineering by executing the SEI’s technical strategy to deliver tangible results.

Researchers highlight methods, prototypes, and tools aimed at the most important problems facing the DoD, industry, and academia, including AI engineering, computing at the tactical edge, threat hunting, continuous integration/continuous delivery, and machine learning trustworthiness.

Learn how our researchers' work in areas such as model-based systems engineering, DevSecOps, automated design conformance, software/cyber/AI integration, and AI network defense—to name a few—has produced value for the U.S. Department of Defense (DoD) and advanced the state of the practice.

Day 1 Monday, November, 14, 2022

AI Engineering in an Uncertain World

Principal Investigator

Dr. Eric Heim

The goal of this research is to detect machine learning model uncertainty and mitigate its effects on the quality of model inference.

Read More

Portable High-Performance Inference on the Tactical Edge (PHITE)

Principal Investigator

Dr. Scott McMillan

PHITE applies performance engineering processes to the analysis of existing open source ML frameworks for embedded systems to inform the development and optimization of a portable software library to boost performance for ML applications across a range of embedded devices.

Read More

Automating Mismatch Detection and Testing in ML Systems

Principal Investigator

Dr. Grace Lewis

We are developing a suite of tools to automate ML mismatch detection and demonstrate how to extend descriptors to support testing of ML-enabled systems.

Read More

A Machine Learning Pipeline for Deepfake Detection

Our deepfake detection prototype framework can ingest modes of data and detect at least three types of AI artifacts for each mode with at least 85% accuracy.

Read More

AI Evaluation Methodology for Defensive Cyber Operator Tools

Principal Investigator

Dr. Shing-hon Lau

The goal of this project is to develop a methodology for evaluating the capabilities of an AI defense.

Read More
Day 2 Tuesday, November, 15, 2022

Chain Games: Powering Autonomous Threat Hunting

Principal Investigator

Phil Groce

Our algorithms enable fully autonomous threat hunting by modeling threat hunting as a Cyber Camouflage Game (CCG).

Read More

Maturing Assurance Contracts in Model-Based Engineering

Principal Investigator

Dr. Dionisio de Niz

This project is developing a verification technology that can describe and enforce assumptions of 75% more analyses and validate the conformance of 70% more assumptions in a system implementation than the known state of the art.

Read More

Safety Analysis and Fault Detection Isolation and Recovery Synthesis for Time-Sensitive Cyber-Physical Systems

Principal Investigator

Dr. Jerome Hugues

SAFIR addresses safety analysis of time-sensitive cyber-physical systems in both its theoretical and practical dimensions.

Read More
Day 3 Wednesday, November, 16, 2022

Refactoring for Software Isolation

Principal Investigator

James Ivers

Our automated refactoring solution recommends ways to improve the modularity of existing software, significantly increasing the efficiency of software evolution.

Read More

Automated Design Conformance during Continuous Integration

Principal Investigator

Dr. Robert Nord

This project developed an automated conformance checker prototype for use in a continuous integration workflow that detects nonconformances in minutes rather than months or years.

Read More

Automated Continuous Estimation for Pipelines of Pipelines

Principal Investigator

Dr. William Nichols

This project instruments complex software production environments to continuously monitor the DevSecOps pipeline and use that data to continuously update estimates for cost, schedule, and quality.

Read More

Advancing Algorithms for File Deduplication Across Containers

Principal Investigator

Kevin Pitstick

Our automated container image minimization technology combines and improves on two minimization approaches: pruning and deduplication.

Read More

Semantic-Equivalence Checking of Decompiled Binaries

Principal Investigator

Dr. Will Klieber

Our tool can identify which decompiled functions are likely to be semantically equivalent to the original binary function and which are unlikely to be equivalent.

Read More