The Ledger
Paper Finds a Tor Path to Monero Transaction Origins
A paper describes ProxyMark, an attack framework that uses Monero’s Tor peer connections to capture transactions and potentially link them to a source IP address. Its experiments report success under specified conditions, not a measured rate of deanonymization across Monero users.

Up to 11 of 12 connections
Adversarial peers occupied as many as 11 of a Monero node’s 12 outgoing Tor hidden-service connections in tests described in a paper submitted July 8 and last revised October 26. That foothold could help an attacker capture transactions before they enter Monero’s public network and, under additional conditions, connect them to the source node’s IP address.
Source arxiv.org
The paper, “Deanonymizing Monero Transactions in Tor Network,” introduces ProxyMark, a three-stage attack framework developed by Ruisheng Shi, Shihan Zhang, Yulian Ge, Lina Lan, Qingfeng Zhang and Qin Wang. It targets network metadata: the path a transaction takes between nodes, rather than Monero’s cryptographic protections for transaction details.
Monero uses Dandelion++ to obscure where transactions originate. The researchers say Tor changes how that protection works for nodes that send transactions over hidden-service connections. In the protocol behavior they analyze, a node first sends its own transactions to two selected outgoing Tor hidden-service peers, which then relay them to clearnet peers; forwarded transactions follow a different route.
ProxyMark exploits the forwarding split
That difference creates a way to identify an originated transaction. If an attacker controls one of the selected proxy peers, a transaction received from the target over that connection indicates, under the forwarding rules described in the paper, that the target generated it. ProxyMark seeks to put adversarial peers in the target’s connection set, increase their chances of being selected as proxies, and then link a Tor-side identifier to an IP address using traffic watermarking.
In a controlled connection-occupation test, the researchers filled peer lists with adversarial onion addresses. After one or two restarts, those peers held 7 to 11 of 12 outgoing hidden-service connections. In a separate Monero mainnet test, with one adversarial peer among 12, the researchers raised the observed chance that an attacker-controlled peer was selected as a proxy from 15.3% to 35.7%, a gain of 20.4 percentage points.
The authors also tested whether their watermarking method could associate node identifiers with IP addresses. In experiments involving hidden-service targets, it recorded 100% precision and 93.8% recall; for Tor client targets, it recorded 100% precision and 91.4% recall. Those measurements came from controlled tests, including configurations in which the researchers’ relay served as the target’s sole entry guard, rather than from an uncontrolled survey of Monero users.
The attack depends on specific access
ProxyMark is not a claim that every Monero transaction sent over Tor can be traced. The researchers say the attack requires a target that uses Monero over Tor, attacker-operated hidden-service peers capable of influencing its peer lists and receiving its transactions, and an adversarial Tor relay selected as the target’s entry guard for the IP-linking step.
Through experiments on the live Tor network, Monero mainnet, and testnet, we empirically demonstrate the effectiveness of ProxyMark in successfully deanonymizing transactions originating from Monero nodes over Tor.
Ruisheng Shi, Shihan Zhang, Yulian Ge, Lina Lan, Qingfeng Zhang and Qin Wang, authors of the paper, write.
The paper reports component tests across the Monero testnet, mainnet and live Tor network, with some attack stages evaluated separately. Its hidden-service connection-occupation experiment also scaled the adversary setup down by a factor of 10 for resource and ethical reasons. Those boundaries matter when interpreting the results: they demonstrate feasibility under the paper’s conditions, not a measured rate of successful deanonymization across the live Monero network.
Protocol changes could close the route
The researchers identify the forwarding split as the root cause and propose extending Dandelion++ stem forwarding to hidden-service connections. If a peer could receive either an originated or a relayed transaction during the stem phase, receipt alone would no longer identify the sender as the origin.
Other proposed changes address individual steps: limit onion addresses accepted into peer lists, check reported block heights against network data before choosing proxy peers, and set a strict rate for timed-sync messages that attackers use to encode watermarks. The authors say those measures could raise attack costs or reduce success, while changing transaction forwarding would address the underlying signal. The paper notes that an abstract version of the work appeared at a WWW 2024 workshop; its July 2026 version develops the ProxyMark framework and reports the newer experiments.
Sources
- Original source · Preprint arXiv arXiv:2607.07062
Related stories
- Iron Fish Ends Network Support After Merkle-Tree Inflation Bug
- Monero Futures Open Interest Tops $300M as Liquidation Risk Rises
- ZachXBT Links 87% of FreedomXMR Inflows to Illicit Funds
- Andrey Chmora Proposes Stealth ABBE for Private Ethereum Contracts
- Helius puts programmable privacy rings on Solana mainnet