Cyber Security Researcher
- Civil Service
- Part Time
- London
- 65,000 - 145,000
Job Description
Job summary
We are looking for a Cyber Security Researcher to join the Cyber and Autonomous Systems Team (CAST) at the AI Security Institute. CAST works to research and map the evolving frontier of AI capabilities and behaviours in order to inform critical security decisions that reduce loss-of-control risks from frontier AI. The team focuses on preventing harms from high-impact cybersecurity capabilities and highly capable autonomous AI systems.
As a Cyber Security Researcher, you will play a key role in building environments, challenges, and evaluation infrastructure to benchmark the cyber capabilities of AI systems. You will help design rigorous and realistic tests that measure how well frontier AI models perform on real-world cybersecurity tasks, supporting AISIs work to better understand the risks posed by advanced AI systems and inform safety standards and policy decisions.
This role is well suited to someone with strong practical cybersecurity skills, good Python capability, and an interest in applying those skills to high-impact work at the intersection of cyber security and AI safety. You will work closely with research engineers, infrastructure engineers, and machine learning researchers across AISI, and will contribute to the development of first-of-its-kind evaluation infrastructure within government.
Job description
You will work as part of a highly collaborative team to design and build cyber capability evaluations for frontier AI systems. In practice, this will mean:
- Designing cyber ranges and CTF-style challenges to benchmark AI system performance on cyber security tasks
- Building evaluation environments and scaffolding to test frontier models in realistic cyber scenarios
- Equipping evaluation environments with relevant tools, such as network packet capture utilities, penetration testing frameworks, and reverse engineering or disassembly tools
- Designing metrics and supporting the interpretation of results from cyber capability evaluations
- Working alongside engineers to ensure evaluation environments and infrastructure are robust, reliable, and scalable
- Contributing to the continuous improvement of automated grading, scoring, and evaluation methodologies
- Helping integrate new cyber ranges and security testing environments into AISIs evaluation pipeline
- Writing reports, papers, blogs, and other outputs to communicate findings to technical and non-technical stakeholders
- Keeping up to date with relevant developments in cyber security research and practice
- Contributing to AISIs broader understanding of AI cyber risks and the implications of increasingly capable frontier AI systems
Person specification
Essential Criteria:
- Strong Python skills, including experience writing scripts for automation or security tooling
- Practical experience in at least one of the following areas of cyber security: Penetration testing Cyber range design Competing in or designing CTFs
- Developing automated security testing tools
- Bug bounties, vulnerability research, or exploit discovery and patching
- Experience designing, building, or operating technical environments for cyber security testing, experimentation, or evaluation
- Ability to contribute to the design of realistic cyber capability evaluations, including tasks, environments, tooling, and scoring approaches
- Strong problem-solving skills and the ability to work effectively in technically complex and fast-moving environments
- Ability to communicate technical findings clearly to a range of technical and non-technical audiences
- Ability to work collaboratively with engineers, researchers, and other specialists in a multidisciplinary team
- Strong interest in improving the safety and security of AI systems
Desirable Skills:
- Familiarity with virtualisation technologies such as Proxmox VE and infrastructure-as-code approaches for creating reproducible test environments
- Familiarity with cyber security tools such as network packet capture utilities, penetration testing frameworks, and reverse engineering or disassembly tools
- Experience building or measuring the impact of automation tools on cyber red-teaming workflows
- Experience developing robust and scalable technical infrastructure for testing or evaluation
- Experience writing reports, technical blogs, or research outputs for internal or external audiences
- Active engagement with the cyber security community and evidence of keeping up to date with new research and emerging techniques
- Interest in, or experience related to, AI safety, frontier AI evaluation, or the security implications of advanced AI systems