p2psec
p2psec copied to clipboard
gpestana
ETH privacy
Blockchain is Watching You: Profiling and Deanonymizing Ethereum Users
- Account-based model is inferior than UTXO model from a privacy perspective; “quasi-identifiers” to tag users based on account addresses (user profiling based on quasi-identifiers);
- Study of the Tornado Cash coin mixer privacy based on strong heuristics that decrease the privacy guarantees of non-custodial mixers on Ethereum;
- Variant of Danaan-gift fingerprint attack for Ethereum;
- Authors use node embedding methods to cluster Eth addresses for user profiling in Ethereum
- Authors collected Ethereum addresses and respective links to users based on data from twitter accounts, tornado cash, humanity-dao; From the 4259 addresses collected, they identified 1,155,188 transactions (sent or received) during 5y.
- Exact identification of accounts pairs/users is not a goal of the paper; instead, the goal is to rank plausible deanonymization candidates and with that reduce the k-anonymity of Ethereum accounts.
Problem 1: In Ethereum, native transactions can only move funds from a single sender and a single receiver, with the change being stored in the sender account. Subsequent transactions will re-use the account that received the unspent amount. Account-based model relies on address-reuse on the protocol level.
Proposed solutions:
-
Coin Mixers:
- M ̈obius: Trustless tumbling for transaction privacy
- Mixeth: efficient, trustless coin mixing service for ethereum
- Sharelock: Mixing for cryptocurrencies from multiparty ecdsa
- Tornado Cash
-
Confidential transactions
- AZTEC
- Pgc: Pretty good decentralized confidential payment system with auditability
- Zether: Towards privacy in a smart contract world
Deanonymization vectors:
- Pairing Ethereum accounts from the same user (Section 6)
- Tornado Cash deposit and withdrawals pairs (Section 7) F- ingerprint accounts through Danaan-gift variant (Section 8)
Section 6: Pairing Ethereum accounts from the same user
3 quasi-identifiers user to link accounts from the same user: Active time of the day Gas price selection Location in the Ethereum transaction graph
Evaluation: Given an Ethereum address, order remaining addresses by their Euclidean distance;
Section 7: Tornado Cash deposit and withdrawals pairs
Section 8: Fingerprint accounts through Danaan-gift variant
Conclusions
Actionable insights / open questions
- “... users should avoid sensitive activities on addresses easily linkable to their public identities, such as ENS name or their Twitter handle.” → due to the possibility to link ENS names to which services/service categories have been used over time (e.g. adult/gambling/DeFi, etc..)
- Different wallet softwares use different methods to compute suggested gas prices. Can we fingerprint a wallet software? How to avoid wallet fingerprinting?
- Network-level privacy -- there are several studies showing how wallet privacy is lost when users interact with full nodes or wallet providers. How can the user protect against broadcast and network-level privacy attacks?
- How may browser and mobile wallets affect privacy? (see paper 3. below) What can be done to prevent that?
- Anonymous transaction relayers?
