Alexis Carlier, Zainab Ali Majid
Building realistic defensive cybersecurity benchmarks. Asymmetric Security responds to real cyber incidents and therefore holds data not available in the public domain. We would like to work with MATS scholars to build realistic benchmarks grounded in these real cyber incidents.
Stream overview
Building Realistic Defensive Cybersecurity Benchmarks
Existing cybersecurity benchmarks lack realism, rarely testing how models behave in realistic security scenarios. This is especially challenging in cybersecurity because most relevant data is private.
Asymmetric Security responds to real cyber incidents and therefore holds data not available in the public domain. We would like to work with MATS scholars to build realistic benchmarks grounded in these real cyber incidents.
Mentors
Alexis is the co-founder and CEO of Asymmetric Security. He was previously an AI security fellow at RAND and part of the founding team of GovAI.
Zainab is the co-founder of Asymmetric Security. She was previously a cybersecurity analyst at Stroz Friedberg, where she investigated some of the largest cybersecurity breaches of the past decade (e.g., Cambridge Analytica). She has also published at NeurIPS on AI cybersecurity evaluations. Zainab holds a master’s degree in Physics from Oxford University.
Mentorship style
1 hour weekly meetings by default for high-level guidance. We will respond within a day to async communication.
Representative papers
https://www.irregular.com/publications/cyscenariobench
https://deepmind.google/blog/evaluating-potential-cybersecurity-threats-of-advanced-ai/
https://docs.google.com/presentation/d/1D1gWFuT6AT3kLOqM1xl5YHKPvAhJh-VW/edit?slide=id.p1#slide=id.p1
Scholars we are looking for
Essential:
- Experience implementing AI model evaluations.
Preferred:
- At least one year of professional software engineering experience.
- Strong interest in AI cybersecurity.
Scholars can collaborate with other MATS scholars and can find collaborators on their own. Asymmetric Security staff may also engage deeply.
Project selection
We will assign the project direction; scholars will have significant tactical freedom.
Community at MATS
MATS Research phase provides scholars with a community of peers.
.webp)
During the Research phase, scholars work out of a shared office, have shared housing, and are supported by a full-time Community Manager.
Working in a community of independent researchers gives scholars easy access to future collaborators, a deeper understanding of other alignment agendas, and a social network in the alignment community.
Previous MATS cohorts included regular lightning talks, scholar-led study groups on mechanistic interpretability and linear algebra, and hackathons. Other impromptu office events included group-jailbreaking Bing chat and exchanging hundreds of anonymous compliment notes. Scholars organized social activities outside of work, including road trips to Yosemite, visits to San Francisco, and joining ACX meetups.