Clarification Request - White Paper - Mixing Nodes

[foundingnimo]

Page 33 In the caption for figure 11 claims that a batch of messages to be shuffled is re-shuffled and re-encrypted by multiple nodes in the mix network.

• The rest of the text in the page does not repeat this claim (so it's only in the caption of the figure)
• I think an explanation of why this is done (if it is done) needs to be provided. As it currently reads, there is no good reason to run the same operation through multiple mix-nodes, as a single mix-node should suffice to re-encrypt and shuffle the vote, and provide a ZKP.
• If every mix node handles every transaction- wouldn't there be a possibility for a rogue node to affect all of the votes (instead of affecting only a portion of them)?

[jeffreypstern]

You are correct in saying that a single mix-node should suffice to re-encrypt and shuffle (with corresponding ZKP), assuming one honest mixing node, for complete anonymization. However, by using multiple mix nodes, we prevent a single point of trust.

The purpose of the mixnet here is to provide full anonymity to the voter. If the Authentication servers were to be compromised, one would theoretically be able to tie the voter to their encrypted vote through their pseudonym. Anonymity is provided by performing the following operations:

• Each mix node shuffles and re-encrypts the votes, providing a proof of valid mix.
• Providing a proof of valid mix mathematically guarantees that the post-shuffle votes are equivalent to pre-shuffle.
• While the mix node cannot change the value of the votes themselves ( because it must provide a proof of valid shuffle ), it can potentially store the order of the shuffle to break the anonymity of the voter (hence by using multiple nodes, we prevent a single point of trust)

[foundingnimo]

Ok - I think this should be explained somewhat better in the paper.

[jeffreypstern]

Added in this description to the paper.