Lingwei Chen
Rochester Institute of Technology, NY·Computer Science
Machine learning can help spot cyberattacks across the huge and growing number of internet-connected devices. These detectors usually need large sets of attack examples that someone has labeled by hand.
- Fields
Computer Science
Engineering
- Topics
cybersecurity
intrusion detection
attack detection
- Grants
- 1 known grant, through Feb 2027
What they work on
- Repurpose pretrained transformer models for cyberattack detection through cross-modal adversarial reprogramming, which adds patch-level changes to inputs so few new parameters are needed when labeled data are scarce.
- Develop scalable semi-supervised learning that uses consistency and contrastive regularization to improve pseudo-labeling of unlabeled attack data and reduce label bias.
- Extend the reprogramming and semi-supervised pipelines to combine labeled and unlabeled data for more effective, low-cost security attack detection, released as open-source systems.
More about the research
Such data are expensive, slow to collect, and often lopsided. Some attack types appear many times, while others barely appear at all, so the data may not reflect real-world attacks. This project asks how to build accurate attack detectors from limited or biased data at low cost.
Why it matters
Many organizations cannot afford large, carefully labeled collections of attack data. Detectors that learn well from few or imbalanced examples could make machine-learning-based cyber defense cheaper and more practical. Open-source tools from the project could help security practitioners deploy these methods to protect networks and connected devices.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Philip Yu
University of Illinois Chicago, IL·Computer Science
Integrate open-source anomaly detection tools for tabular, graph and time-series data into a single OpenAD ecosystem with a standardized development environment for building and evaluating models.
Hamed Tabkhi
University of North Carolina at Charlotte, NC·Computer Science
Develop a semi-supervised transformer model that detects unusual or threatening behavior in real time from surveillance video. Use identity-neutral visual features so that no personally identifiable information is needed.
Carl Vondrick
Columbia University, NY·Computer Science·Early-career
Build open-world recognition models for satellite, aerial and ground imagery from vision-language models without language annotations, using cross-modal contrastive learning and reasoning about change over time.
The school
Rochester Institute of Technology: 9 labs on PI Pages, $78M a year on research, 65 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.