Evoting using collaborative clustering

1
Evoting using collaborative clustering

• Justin Gray
• Osama Khaleel
• Joey LaConte
• Frank Watson

2
Overview

• Assumptions / Introduction
• Detailed layout of the system
• Security

3
Assumptions

• The computer has a static IP address and is
  powered on all the time.
• The user has administrative privileges to the
  computer.
• The IT department has one unused server also with
  a static IP address.

4
What were planning to do

• Were going to make all the faculty machines part
  of an overlay network. The machines communicate
  with each other and randomly select (3-4
  computers) to act as counters.
• When someone casts a vote, the encrypted data is
  sent counters and tallied.
• Well use PKI to securely communicate the between
  the different counters.
• When the election is over the counters broadcast
  the information to the other machines in the
  network.

5
What inspired this design

• Byzantine fault tolerance this idea is used in
  failsafe systems (such as aircraft) where there
  is redundancy to make sure the decision is
  correct.
• We have three or four hidden and random counters
  which are sent data. If one of them miscounts or
  is compromised, the other counters in the cluster
  will help validate the vote.

6
Adding to the Voting Cluster
1. System admin adds the users credentials to
the server.
2. Software is installed in the facultys
members computer and the computer sends the
server its public key.
4. A signed message is sent to all the computers
in the network a new computer is added to the
cluster and updates them with new computers
certificate.
3. The server replies back with a signed
certificate using the clients public key.
7
Adding a machine to the cluster (cont.)

• The system will have a server that will be
  responsible for adding other machines to the
  voting cluster, plus providing CA services.
• The server will be pre-configured with a list of
  legitimate IP addresses that can join the
  cluster.
• Once the service starts on the client side, it
  sends a request (including IP, machine name, MAC,
  ) to the server to add itself. The server checks
  the IP (or maybe the MAC) against the allowed
  list, and records other information.

8
Certificates distribution

• The next step the client should do is to generate
  a public/private key pair.
• We provide the servers public key integrated
  (hard-coded) in the client software.
• So, the client will use the servers PK to
  encrypt his/her PK, and then send it to the
  server.
• The server decrypts it using its private key,
  generates a certificate, and broadcast it to all
  joined machines so each machine can sign a ballot
  and authenticate other machines.

9
Voting Process

1. Ballot Creation

• Distribute Ballot

• Randomly select counters

1. User Votes

• Submit completed vote to counters
• Generate receipt

10
Voting Process

1. Counting Votes

• Heartbeat / NTP

• Tally results

1. Recount Votes

• May occurs if discrepancy
• Collect printed receipts

11
Security

• PKI
• This evoting system makes heavy use of PKI,
  inasmuch as every member of the cluster has their
  own private key, they also have the public key
  for every other member of the cluster. This
  allows all the nodes to communicate with each
  other securely.
• By securely, we mean that every voter is able to
  know that the message they received came from
  whom it was supposed to come from, and they can
  rest assured that only they were able to read the
  message destined for them.

12
Security

• Encryption
• Encryption is used primarily at two levels in
  this system
• Message level
• All messages are signed and encrypted. This
  includes messages related to joining the cluster,
  and messages related to voting.
• Log level
• All votes are stored in an encrypted form in a
  file that can be retrieved in case of
  catastrophic failure. Votes are kept in a hash
  form to ensure that the voter who casts a vote
  remains anonymous.

13
Security

• Distributed Voting
• The distributed nature of this system has its own
  security benefits.
• Vote tampering
• Since the vote counters are chosen at random from
  within the cluster, it would be difficult to find
  and attack a vote counter.