Skip to content

Research data ​

Research snapshot v0.2.0 contains ten enriched Ethereum incidents and 55 component metadata snapshots. Nine incidents have proposed defect source. Components include proxies and dependencies; they are not 55 vulnerable victims.

See three real examples on the landing page, including an unresolved source gap. The downloadable preview contains metadata and source fingerprints, not contract code.

Inspect the snapshot

Newer catalog entries ​

The separate pinned DeFiHackLabs catalog contains 357 Ethereum entries, including 58 dated 2025 and 59 dated 2026, as recorded at upstream commit 23c7bd83ccb4524972a9493617fcc8764ab6269b on 21 October 2026. These are catalog entries, not 357 independently verified hacks or members of the ten-incident enriched snapshot. The landing page has separate 2026, 2025 and enriched views. Catalog dates are reported incident dates; completeness, source coverage and human adjudication are not established by catalog inclusion.

Inspect the catalog

What is published ​

Addresses, evidence links, hashes and fetch instructions. Contract source, PoCs and article/post bodies are not redistributed. Snapshots are immutable; new research needs a new version.

Labels remain AI proposals. Human adjudication, earliest disclosure, family separation and training/evaluation eligibility remain separate gates.

Verify an export ​

sh
python3 -I m0/scripts/discovery/export_research.py \
  --verify datasets/research/v0.2.0

Practical trials ​

The workflow has been tried on a no-change OpenZeppelin beacon review and a known historical ERC165Checker fix proposal. The latter preserved a baseline and comparison, but compilation and runtime/resource regressions were not run because no container runtime was available. Neither trial was a blind detection benchmark.

Runtime validation limits

Early research tooling. A valid report is not a security certificate.