
- Curating timing, gas and error-code traces as LeakDetect.ai training data
- Privacy-preserving log packaging on IPFS with IPLD content addressing
- Threat taxonomy for observability-derived leakage in Ethereum clients

A community-led Ethereum security research and education initiative focused on identifying attack paths across wallets, dApps, AI agents, developer tools, and underlying P2P networking. Privacy and metadata protection support this core security mission.
LeakDetect.ai is supported through the EF DAO Fund via Giveth, has been invited to the 1 Trillion Security initiative of Ethereum, and is supported by Wintermute, with advice and technical input from members of the libp2p, IPFS and Filecoin communities.
Rather than treating smart contracts, wallets, agents, and P2P networks as isolated layers, the hub examines how weaknesses and attacks can cross them, and how security measurement, authenticated communication, and resilient infrastructure reduce risk.
LeakDetect.ai is the continuous research and measurement thread across all four days, connecting individual sessions into one broader investigation of Ethereum's attack surface.
Establishes a common security framework for thinking about Ethereum beyond smart contracts.
Can we trust the communication channels through which wallets, agents, dApps and developer tools exchange security-sensitive information?
The security layer that is often invisible to application developers: Ethereum's underlying P2P networking infrastructure.
How should Ethereum's networking layer evolve for a post-quantum future?
User ↓ Wallet ↓ dApp / Agent ↓ RPC / API ↓ Ethereum L1 / L2 ↓ libp2p / P2P Network ↓ Peers / Relays / Observers
| Day | Theme | LeakDetect.ai focus | Practical output |
|---|---|---|---|
| Day 1 | Cross-Layer Security Threat Modeling | Zcash/Tornado Cash datasets, metadata exposure, threat modeling | Cross-layer threat models |
| Day 2 | Secure Wallet & Agent Communication | Communication-channel leakage and authentication | Secure communication models |
| Day 3 | P2P Resilience | Luminar, Sybil, eclipse and partition detection | P2P measurement experiments |
| Day 4 | Post-Quantum Ethereum | PQ signatures, PKIX, L1/L2 and PQ light clients | PQ networking research roadmap |
Two community workshops delivered across 29 centers with roughly 750 attendees, covering protocol analysis, interop testing and observability-driven research.






Four additional two-hour workshops delivered in collaboration with MeitY's India AI Mission as part of Road to Devcon. Every session keeps LeakDetect.ai as the central theme: building, curating and annotating the datasets that power leakage detection across Ethereum.





LeakDetect.ai brings together security research that is often fragmented across Ethereum's application, wallet, agent and networking layers.
The hub gives Devcon attendees an opportunity to move beyond purely smart-contract-centric security and examine how wallet and agent weaknesses, insecure communication, P2P attacks, and cryptographic assumptions interact across the complete Ethereum stack. By combining real-world datasets, hands-on threat modeling, network measurement with Luminar, libp2p engineering and post-quantum research, the hub creates a practical environment where researchers and builders can identify vulnerabilities, test defensive approaches and develop ideas that continue beyond Devcon.
