AI-based Hardware Attacks (AHA!) Challenge
Use LLMs to insert stealthy Trojans and to hunt them down!
The AI-based Hardware Attacks (AHA!) Challenge is a hands-on Hardware Hacking Competition at IEEE HOST 2026 that explores the emerging role of Large Language Models (LLMs) in both attacking and defending hardware systems.
This competition follows a red-team / blue-team format centered on hardware Trojans:
- Red Teams will use LLMs to design and insert stealthy Trojans into small benchmark circuits.
- Blue Teams will leverage LLMs and analysis tools to detect, localize, and reason about these malicious modifications.
Competition Guidelines:
Participating teams will be tasked with using generative AI tools to first insert and exploit hardware security vulnerabilities and Trojans into a large open-source hardware project (to be announced later), then detect and suggest repairs for those vulnerabilities from other teams. This will be done in two phases, one red team (attacking) phase and one blue team (detecting/defending) phase. Each phase will be given one week, both prior to HOST 2026, with in-person live presentations and demos to take place for the judges at HOST. Full guidelines are given below:
Teams:
Teams must consist of up to 4 students and must have one advisor
Phase 1 (Red Team):
Attack challenge specifics (attack surface and any related infrastructure) will be given at the start of the phase and Phase 1 will last one week.
All Trojans must be made by the team’s generative AI system.
Teams can manually copy/paste the Trojans into the HDL if necessary, but the Trojan cannot be modified and any additional changes to the surrounding HDL must be dictated by the AI tool.
The attack phase will be judged based on both the individual Trojans inserted as well as the capabilities of the generative AI tool/framework to design, create, and inject these Trojans.
Phase Deliverables:
Full source code for the team’s Trojan insertion framework, made open-source.
Complete source code for each file with a Trojan or part of a Trojan inserted.
A brief README detailing the design and operation of the team’s framework, as well as explanations of each Trojan.
Logs for all interactions with the generative AI framework including direct prompts, thought processes (for reasoning models), inter-agent communications, etc.
Phase 2 (Blue Team):
The detection/defense phase will focus on detecting and offering corrections for Trojans inserted by the other teams in Phase 1.
Teams will receive one copy of the target hardware design for each other team containing that team’s Trojans.
All Trojan detection must be done completely using a generative AI tool or framework.
Phase Deliverables:
Full source code for the team’s Trojan detection framework, made open-source.
A README detailing the methods used by the tool or framework to detect Trojans.
Logs for all interactions with the generative AI framework including direct prompts, thought processes (for reasoning models), inter-agent communications, etc.
Details on each Trojan detected across the provided deliverables from Phase 1.
Final Presentation:
Presentation length TBD.
This presentation must include a discussion of the methods used to utilize generative AI
The presentation must include details on the Trojans created in Phase 1
The presentation must include details on the Trojans detected in Phase 2
Final Deliverables:
The final presentation
A final document detailing the use of AI throughout both phases, including any design and usage docs for tools or frameworks
All AI logs for both phases
Competition Webinars:
March 20th at 11AM EDT via Zoom (https://nyu.zoom.us/j/91912162661)
April 10th at 12:00PM EDT via Zoom (https://nyu.zoom.us/j/93391346633)
Timeline:
April 8: Registration deadline
April 10: Demo webinar, showing a basic example of using an LLM to add and detect a Trojan
April 12 – 18: Phase 1 of the competition (red team)
April 26 – May 2: Phase 2 (blue team)
May 6: Demos/presentations at HOST & announcement of winners at the award ceremony
Registration is now open!
Interested participants are invited to sign up using the registration form below. More details about rules, benchmarks, evaluation criteria, and timelines will be released soon.
Link to the Registration Form: https://forms.gle/vMmvr9WCqgkrUnUo8
Link to the Discord Channel: https://discord.gg/8Ruze7SEbN