Can Tor Directory Authorities Deanonymize Users?

A single compromised Tor directory authority cannot directly deanonymize users due to Tor’s distributed consensus architecture. Directory authorities maintain the official list of active relays on the network, but no individual authority has access to user traffic or circuit encryption keys. While a solitary compromised authority poses minimal direct risk, a coordinated compromise of the majority of authorities could allow an adversary to manipulate the network consensus, facilitating targeted deanonymization attacks.

What Tor Directory Authorities Do

Tor relies on a small group of trusted servers known as Directory Authorities (DirAuths), currently numbering nine located around the world. These servers continuously test and monitor the Tor network to determine which relays are active, stable, and secure. Once an hour, they vote on the network’s state to produce a cryptographic document called the “consensus.” Tor clients download this consensus to select their three-hop circuits: the Guard node, the Middle relay, and the Exit node.

Why a Single Compromised Authority Cannot Deanonymize You

Directory authorities do not route user traffic or manage user encryption keys. Consequently, a compromised authority cannot read your data or map your IP address to your destination.

Additionally, the consensus generation relies on a majority voting mechanism. For a relay to be listed or given specific status flags (such as the Guard or Exit flag), a majority of active directory authorities must agree. A single rogue authority cannot force malicious relays onto the network or strip legitimate relays of their flags because the remaining honest authorities will outvote it.

Risks from a Single Compromised Authority

While direct deanonymization is not possible with one compromised authority, an attacker could still attempt lower-level disruptions: * Early Detection of Network Changes: The attacker might gain insight into unannounced relays or early network shifts. * Biased Individual Metrics: The rogue authority could submit skewed measurements regarding relay bandwidth or uptime, attempting to slightly influence the consensus weights, though the impact remains diluted by the honest majority. * Targeted Information Gathering: If a client directly queries the compromised authority for directory information (via fallback mechanisms), the authority learns the client’s IP address, but still cannot trace the client’s destination or decrypt their traffic.

The Threshold for Deanonymization: Majority Compromise

If an adversary successfully compromises a majority of the directory authorities (e.g., at least 5 out of 9), the risk profile changes drastically. With majority control, the attacker can: 1. Forge the Consensus: Inject hundreds of malicious Guard and Exit relays into the consensus while assigning them artificially high bandwidth weights. 2. Execute Sybil Attacks: Remove legitimate relays to force user traffic through attacker-controlled nodes. 3. Perform Traffic Correlation: If an attacker controls both the Guard node (which knows the user’s IP) and the Exit node (which knows the user’s destination), they can correlate traffic timing and packet sizes to deanonymize the user.

Tor’s directory authority design intentionally separates administrative consensus from data routing. Deanonymization requires end-to-end traffic correlation across a user’s circuit, a feat that a single compromised directory authority cannot accomplish alone.