axiom
ColdTrust
AI Brain
OLX India CLI
ForensicToolW3
Love LaceGround truth, not scraped logs
Every attack is a reproducible simulation with the source code and validation files beside it, so you know exactly which transactions are malicious.
AttackLogs
EVM attack simulation dataset — 36 documented smart-contract exploits running on a local chain, producing forensic-grade NDJSON that matches a published schema.
The problem
Detection tools get benchmarked against public logs scraped after the fact. Nobody knows which transactions were the attack, so the score means nothing.
The approach
Generate the ground truth instead: run the real exploit, keep the source and the expected signals beside the trace.
What is there
36 documented exploits reproduced as Foundry simulations, emitting forensic-grade NDJSON across seven entity files, with the attacker-controlled addresses marked.
Stack
5
- EVM
- dataset
- Solidity
- Foundry
- forensics
Capabilities
2
Built to benchmark against
Train a detector or test a rule-based engine against known attack classes without touching mainnet.
