THREATBYTESTM research
Research aim
To design and develop an agentic, cognitive, and reasoning-driven SI framework capable of predicting, detecting, analysing, and autonomously mitigating novel malicious software, using historical, simulated, and community-submitted malware data for proactive intelligence and pre-execution prevention.
Strategic shift01
The doctoral programme redefines defensive strategy from reactive detection to predictive malware intelligence. Outcomes target autonomous threat intelligence that can withstand adaptive and SI-assisted malware generations while preserving governance, explainability, and human oversight for high-impact actions.
Scope02
In scope:
- Malicious executables and scripts drawn from historical corpora, sandbox execution, synthetic scenarios, and governed community submissions.
- Prediction, detection, analysis, and mitigation inside controlled research environments.
Out of scope:
- Building or releasing offensive tooling.
- Autonomous action on networks the programme does not control or have consent to protect.
- Attributing malware to named individuals.
What success looks like03
- Higher pre-execution detection than reactive baselines on held-out families and later time periods, at a comparable or lower false positive rate.
- Mitigation proposals that human reviewers judge correct and proportionate.
- Explanations that analysts can check against the evidence.
- Datasets, configurations, and models versioned well enough for others to reproduce the findings.