0bin

0bin is client side encrypted pastebin that can run without a database.

It allows anybody to host a pastebin while welcoming any type of content to be pasted in it. The idea is that one can (probably...) not be legally entitled to moderate the pastebin content (http://www.zdnet.com/blog/security/pastebin-to-hunt-for-hacker-pastes-anonymous-cries-censorship/11336) as one has no way to decrypt it.

It's an Python implementation of the zerobin project (https://github.com/sebsauvage/ZeroBin/).

How it works

When creating the paste:

• the browser generate a random key;
• the pasted content is encrypted with this key using AES256;
• the encrypted pasted content is sent to the server;
• the browser receives the paste URL and add the key in the URL hash.

When reading the paste:

• the browser makes the GET request to the paste URL;
• because the key is in the hash, the key is not part of the request;
• browser gets the enrypted content et decrypt it using the key;
• the pasted content is decrypted and code is colored.

Key points:

• because the key is in the hash, the key is never sent to the server;
• therefor it won't appear in the server logs;
• all operations, including code coloration, must happens on the client;
• the server is no more than a fancy recipient.

Technologies used

• Python
• The Bottle microframework
• SJCL
• jQuery
• Bootstrap
• Cherrypy (server only)

Known issues

• 0bin use several HTML5/CSS3 features that are not widely supported. In that case we handle the degradation as gracefully as we can.
• The "copy to clipboard" feature is buggy under linux. It's flash, so we won't fix it. Better wait for the HTML5 clipboard API to be implemented in major browsers.
• The pasted content size limit check is not accurate. It's just a safety net, so we thinks it's ok.
• Some url shorteners and other services storing URLs break the encryption key. We will sanitize the URL as much as we can, but there is a limit to what we can do.

What does 0bin not implement?

• Request throttling. It would be inefficient to do it at the app level, and web servers have robust implementations.
• Hash collision: the ratio "probability it happens/consequence seriousness" is not worth it (http://stackoverflow.com/questions/201705/how-many-random-elements-before-md5-produces-collisions)
• Comments: for now. It's no the todo list.