AI

AMAB allows users to study adversarial manoeuvre attacks, i.e., malicious driving manoeuvres designed to influence a target driving policy through ordinary vehicle behaviour. More specifically, it supports the workflow needed to (1) define a scenario, (2) train a driving policy used as the target victim, (3) train a reinforcement-learning attacker, and (4) evaluate the attack effectiveness. The platform is implemented in MetaDrive and uses attackers based on Stable-Baselines3. The related VehicleSec ‘26 work is A Benchmark for Security Evaluation of Adversarial Manoeuvre Attacks in Autonomous Driving.

10 Aug 2026

BigEye is an autonomous fuzzing campaign manager that turns a source repository into a continuously managed, inspectable security campaign. BigEye is not a new fuzzing engine. It manages AFL++ and libFuzzer for maintainers, security researchers, product-security engineers and application-security teams that need useful fuzzing to continue after the first harness has been generated. BigEye was submitted to OpenAI Build Week. At the time of writing, it is under evaluation.

21 Jul 2026

This project focuses on advancing automated security testing of GraphQL APIs through innovative research and tooling. We have developed and released two major research contributions that enhance GraphQL testing capabilities through different approaches. Research Contributions Wendigo: Deep Reinforcement Learning for Denial-of-Service Query Discovery in GraphQL Wendigo is a black-box Deep Reinforcement Learning approach that discovers Denial-of-Service exploitable queries against GraphQL applications. Using only the GraphQL schema, Wendigo can discover queries capable of performing DoS attacks with just two requests per hour, as opposed to the high volume required by traditional attacks. BenGQL: An Extensible Benchmarking Framework for Automated GraphQL Testing (ASE 2025) BenGQL is an extensible benchmarking framework containing 23 representative open-source GraphQL server applications. This framework enables rigorous evaluation of automated testing tools across different GraphQL engines and schema complexities. Ongoing Research We are currently exploring advanced AI techniques, including Large Language Models, to further enhance GraphQL security testing capabilities. This research aims to develop more sophisticated and context-aware testing approaches. Collaboration Interested in collaborating on GraphQL security research? We welcome partnerships with researchers and industry professionals working on API security, automated testing, and AI-driven security tools.

15 Sep 2025

PolyMirror.AI is a Web3 application that enables paying for AI services using cryptocurrency (POL) through EIP-712 vouchers. The project received Honourable Mention in the “Polygon Track” at the Vibe Coding Hack by Encode.

22 Jun 2025

TonAI Stark is an AI-powered DeFi assistant for StarkNet that won 2nd Place in the “Starkware AI x DeFi” Track at the Encode AI London 2025 Hackathon. Think JARVIS meets Web3 - an intelligent assistant that simplifies crypto interactions with clarity, confidence, and a touch of sarcasm.

15 Feb 2025

Flare-FL is a decentralised Federated Learning (FL) framework built on the Flare blockchain, developed for ETHOxford 2025. The project won Pool Prize in the “Flare: Enshrined Data Protocols” Track and Nerdo Awards in the “DeSci World” Track for being the “most likely to disrupt” project.

9 Feb 2025

Proteus is an LLM-powered lending agent for the Aave V3 protocol designed to automate and optimise lending strategies. The project won 1st Place in the RNDM Agent Track at the Encode London 2024 Hackathon.

25 Oct 2024