AI Safety

Internet anonymity without Tor

A new research paper outlines a radical, bandwidth-intensive alternative to Tor for extreme threat models.

Deep Dive

Researcher Samuel Shadrach has published a detailed analysis proposing a radical alternative to Tor for achieving intelligence-agency-resistant internet anonymity. The paper, titled 'Internet anonymity without Tor,' starts from the premise that physical network infrastructure—specifically fiber optic cables—can be wiretapped by powerful adversaries, allowing them to collect metadata (sender, receiver, timestamps) even if message content is encrypted via PGP. This metadata correlation is a fundamental weakness of standard communication methods against a 'global adversary' with infrastructure access.

The proposed solution is a 'send to all' broadcast model. Instead of sending a message to a single recipient, each user would broadcast a fixed daily payload (e.g., 1 MB of PGP-encrypted data) to every other user on the network at a synchronized time. Recipients would download the entire global message dump to find messages intended for them, using hidden recipient encryption to obscure the target. This floods the adversary with data, making it computationally infeasible to correlate senders and receivers based on timing. However, the scheme is massively bandwidth-intensive: Shadrach calculates that with 100 million users each on a 1 Gbps unmetered connection, each user would have only about 105 KB per day for their actual payload, with the rest being junk data from others.

The paper is a theoretical exercise for extreme threat models, acknowledging Tor currently has no publicly known deanonymization attacks. Shadrach himself notes he wrote it before focusing on AI safety and now leans toward transparency over privacy. The work highlights the severe physical and economic constraints of achieving true anonymity against nation-state actors who can monitor the internet's backbone, presenting a fascinating, if impractical, cryptographic thought experiment.

Key Points
  • Proposes a 'send to all' broadcast model where users send daily encrypted payloads to everyone, masking metadata from fiber optic taps.
  • Calculates severe bandwidth limits: with 100M users on 1 Gbps, each gets only ~105 KB/day for actual message payload.
  • Aims to defeat intelligence agencies that can wiretap global infrastructure and correlate sender/receiver timestamps, a key Tor vulnerability.

Why It Matters

Highlights the extreme difficulty of true anonymity against global surveillance and the physical limits of cryptographic systems.