On 9/23/06, Alison Wheeler <wikimedia(a)alisonwheeler.com> wrote: During the first election I asked this and what I got from the discussion was that this can't be done. The process used for encryption generates random padding so that re-encrypting the exact same message using the same public key will produce a different result every time. My "receipt" did not indicate any information about this random string/padding. Things might have changed, or maybe I was informed incorrectly about this the first time. In any case the message certainly seems to contain more information than just your vote, as I'm pretty much certain that someone else voted exactly the same way as me and yet my encrypted vote is not duplicated in the list of votes. (Doing a google search appears to confirm that we all had the same encryption and signing keys of 0x4E86F78C and 0xA12C1339, respectively.) If you copied your "resulting encrypted version" when you voted, then you can look at [[Special:Boardvote/dump]] to ensure that it hasn't been tampered with *since voting*. Of course this doesn't ensure that your vote wasn't tampered with *at the time of voting*. If what I've said above is correct, the only ways to do that would be to either decrypt your vote with the private key or to obtain the information about the random padding from someone who has access to the private key. That private key is almost surely not going to be released to the public, though it could theoretically be used to spot check certain votes. As for releasing the random padding information to anyone who wants to check their own vote, that's probably possible, assuming there is no information in the raw (padded) unencrypted message which is sensitive. Anthony

Alison Wheeler wrote: The voting system allows for spot-checks as follows: * Download [[Special:Boardvote/dump]], and check that your record is there * Check that your encrypted record has been signed with the appropriate secret key (some votes in the first few hours of voting were signed with the expired 2005 key, but most were signed with the 2006 key) * Contact an election administrator, confirm your identity. Tell them your encrypted record and who you voted for. Ask them to decrypt your election record to ensure that the two match. Spot checks like this provide some assurance against wide-scale falsification of the records. However, the voting system is not perfectly secure. For example, with root access to the servers, you could add false votes for nonexistent people to the dump. This is theoretically detectable, but it would be possible for such things to go unnoticed. Small-scale fraud (small enough to escape random checks) is also possible by compromising the client computer. And since we don't yet use SSL, there is some vulnerability to compromise of the communications channel. I hope I'm not giving people a false sense of security by implementing all this encryption stuff. The goal is only to make detection of attacks easier, or at least theoretically possible. The absolute security of the system still depends on the security of the constitutent parts, namely: the election administrators, the servers, the clients and the network. There is plenty of room for improvement in each department. -- Tim Starling