Check out dense-subset-sum (made in Rust): https://github.com/bc1cindy/dense-subset-sum
Stealth currently focuses on heuristics that detect privacy leaks in normal wallet behavior (reuse, CIOH, change detection, etc.), but it could reach further to ensure user UTXO set safety by evaluating CoinJoin transactions.
dense-subset-sum is a WIP tool that counts (or approximates) how many plausible input output mappings exist for a given CoinJoin. The count is the privacy primitive because many alternative decompositions means good privacy, few mean the CoinJoin barely hides anything.
One way it could plug into stealth engine is as a WEAK_COINJOIN detector (stealth would need a CoinJoin classifier first), flagging txs that look like CoinJoins but whose W(E) is low enough that an observer can narrow the mapping down. It could also work the other way, as a positive signal when scoring downstream UTXOs that came out of a high entropy CoinJoin.
Just a suggestion, figured it was worth surfacing.
Check out
dense-subset-sum(made in Rust): https://github.com/bc1cindy/dense-subset-sumStealth currently focuses on heuristics that detect privacy leaks in normal wallet behavior (reuse, CIOH, change detection, etc.), but it could reach further to ensure user UTXO set safety by evaluating CoinJoin transactions.
dense-subset-sumis a WIP tool that counts (or approximates) how many plausible input output mappings exist for a given CoinJoin. The count is the privacy primitive because many alternative decompositions means good privacy, few mean the CoinJoin barely hides anything.One way it could plug into stealth engine is as a
WEAK_COINJOINdetector (stealth would need a CoinJoin classifier first), flagging txs that look like CoinJoins but whoseW(E)is low enough that an observer can narrow the mapping down. It could also work the other way, as a positive signal when scoring downstream UTXOs that came out of a high entropy CoinJoin.Just a suggestion, figured it was worth surfacing.