Ill Bloom: Dataset Construction Methodology
This document describes how Ill Bloom Exposed Address Set 1 of June 2026 was constructed. It covers the methodology used to identify affected recovery phrases and build the analyzed address set.
Scope
This dataset includes only 12-word BIP39 recovery phrases (128-bit entropy) using the English and selected non-English BIP39 wordlists.
The analysis is intentionally iterative due to computational and data processing constraints. Ill Bloom Exposed Address Set 1 of June 2026 covers a subset of commonly used blockchain networks and derivation paths selected to provide high signal coverage for seed-level detection while keeping the search computationally practical.
Future address sets may expand the supported recovery phrase types, blockchain networks, derivation paths, or search space.
How the dataset was constructed
The Ill Bloom Exposed Address Set 1 is constructed in two stages:
In Stage 1 — Seed discovery, we generate candidate recovery phrases from the vulnerable recovery-phrase generation process and compare their derived addresses against public blockchain datasets built from Google BigQuery. Recovery phrases with at least one matching address proceed to the next stage.
In Stage 2 — Address expansion, we deterministically derive the additional addresses defined in the expansion scope associated with each recovery phrase and add them to the dataset. This expansion is performed regardless of whether those addresses have previous on-chain activity or currently hold funds.
Stage 1 — Seed discovery
For each candidate recovery phrase, the following derivations are evaluated:
| Scheme | Path |
|---|---|
| Ethereum / EVM-compatible networks | m/44'/60'/0'/0/0 |
| Tron | m/44'/195'/0'/0/0 |
| Bitcoin P2PKH | m/44'/0'/0'/0/0 |
| Bitcoin P2WPKH | m/84'/0'/0'/0/0 |
| Bitcoin P2SH-P2WPKH | m/49'/0'/0'/0/0 |
We focused on seed discovery and therefore evaluated only the first address of account 0 (index 0) for each supported derivation path. This approach reduced the computational cost of searching the full candidate recovery phrase space while preserving a strong signal for identifying seeds with observable on-chain activity. Matches count regardless of whether the address currently holds funds.
Note to researchers: The methodology is representative of the discovery process used in this research. The final recovery phrase count reported for this dataset also incorporates recovery phrases identified through additional research efforts.
On-chain activity datasets
Stage 1 uses address datasets built from public Google BigQuery blockchain data.
For this address set, the datasets represent addresses with observable on-chain activity:
- Bitcoin: addresses appearing in transaction outputs. The dataset includes approximately 1.5 billion addresses.
- Ethereum: addresses appearing as transaction senders. The dataset includes approximately 341 million addresses.
- EVM-compatible networks: addresses appearing as transaction senders across Polygon, Arbitrum, Optimism, Avalanche, Fantom, and Cronos. The dataset includes approximately 190 million addresses.
- Tron: addresses identified through USDT (TRC20) transfer activity and balance thresholds. The dataset includes approximately 254 million addresses.
Stage 2 — Address expansion
Once a recovery phrase passes Stage 1, we deterministically derive the additional addresses associated with that recovery phrase.
Unlike Stage 1, this step does not search for on-chain activity. Instead, it reconstructs the set of addresses that the same recovery phrase can generate across the supported derivation paths and blockchain networks.
The resulting address set is derived from approximately 2,805 recovery phrases and is the dataset used by the Is my wallet exposed? tool.
Expansion scope
| Network / Scheme | Path |
|---|---|
| Ethereum / EVM-compatible networks | m/44'/60'/0'/0/{0,1,2} |
| Ethereum (Ledger-style accounts) | m/44'/60'/{1,2}'/0/0 |
| Bitcoin BIP44 | m/44'/0'/{0,1,2}'/0/0 |
| Bitcoin BIP49 | m/49'/0'/{0,1,2}'/0/0 |
| Bitcoin BIP84 | m/84'/0'/{0,1,2}'/0/0 |
| Tron | m/44'/195'/0'/0/0 |
| Rootstock | m/44'/137'/0'/0/0 |
| Cosmos | m/44'/118'/0'/0/0 |
| Solana (Phantom) | m/44'/501'/0'/0' |
| Solana (Solflare) | m/44'/501'/0' |
| Nano | m/44'/165'/0' |
For Bitcoin, we also generate the supported address encodings for the derived key material, including P2PKH (compressed and uncompressed), P2WPKH, and P2SH-P2WPKH.
Dataset limitations
Ill Bloom Address Set 1 of June 2026 represents a snapshot of the analysis performed using the recovery phrase types, blockchain networks, and derivation paths described above.
Coverage depends on the defined search scope and available on-chain datasets. Recovery phrases that do not produce a matching address during Stage 1 discovery are not included in this address set.