On-Demand ONions: utility to create Tor hidden services on-demand
At 31C3, there was discussion of possible censorship mechanisms for Tor. While the context was laudable (preventing Tor from becoming a heaven for child pornography sites), I don't think a mechanism for censorship should ever be built into Tor.
The proposed solution was to distribute a blacklist of .onion addresses (or their hashes, or a bloom filter, or something similar) to relays or clients, opt-in or opt-out. This tool exists as a quick proof that such a mechanism wouldn't work, because blacklists can't work in an environment where address creation is fast, cheap and decentralized whereas maintaining the blacklist and distributing it is a lot more expensive.
The basic idea is as follows:
- A hidden service can have N addresses for "advertising". These addresses are advertised to new users of the service via other channels,
- When an user visits an advertising address, they obtain a new address. This address is generated on-the-fly to point to the actual content. It is never shown to another user, so the only person able to block it is the person who got it. It is impossible to exhaust all the addresses generated this way as it would require exhausting all possible .onion address space (2^80 addresses, about 10^24).
- In case a user's address is blocked and they lose access to the advertising service, they may have the option to generate backup addresses that they can keep secret.
This is just a quick experiment I came up with during the last hours of 31C3. It is not used in production anywhere. But it is usable.
The Dockerfile provided lets you run a tor daemon and odon in a docker container.
$ docker build -t tor .
$ docker run -p 9051 -p 5000 tor
To use odon from a browser, navigate to http://localhost:5000/.
You need to set up your own hidden service to start with. This service will lead to a Python application handler, e.g. using flask. Look at the provided code on how flask is integrated to generate hostnames on demand.
Odon needs to have read access to the Tor config. On many systems this means chmodding a lot of files, or running Odon as root (you probably shouldn't do that). Adapting the existing docker-based solution may be preferrable for your needs. Be aware, however, that Docker doesn't provide a security barrier, and jobs running as root in a docker container may escape the container.