Secure Voting in US Elections

Broadcasting from Colorado on February 21 2025, he tackles the common question of whether blockchain can make the American election system far more secure, answering maybe.

Introduces a paper he finds deeply unfair by famous academics Neha Narula of MIT's Digital Currency Initiative and Ron Rivest, the R in RSA and a Turing Award winner.

The MIT paper concludes blockchain does not solve the fundamental security problem of electronic voting and may create new attack surfaces, which he disagrees with.

Shows a real hybrid system combining paper ballots, blockchain storage, secure digital receipts, biometric verification, and risk audits as the right approach.

States asking for blockchain voting get a no unless identity is included, since nations identify people through breeder documents like driver's licenses, passports, and birth certificates.

Voter ID security depends first on identity security, and breeder documents are notoriously unreliable, easy to counterfeit, and even forgeable by governments abroad.

Points to open standards like the W3C DID v1 decentralized identifier and AnonCreds, which prove a property such as being over 21 without revealing name or age.

Secure identity is only the entry point, and next you need secure registration so no one registers dead or fake people even with unforgeable tallies.

Registration also needs proof of life, proof of location, and proof of good standing, checking residency and felon eligibility against voting criteria.

You then need an unforgeable ballot that cannot omit legitimate candidates, citing Robert Kennedy's ballot access issues and the risk of injected fake paper ballots.

He hates mail in ballots as the least secure method, with no chain of custody, making it the easiest way to sell your vote or have ballots intercepted.

Signature verification is weak, but biometrics like a fingerprint at registration at least prove Bob not Alice voted, though the paper argues phones can be hacked.

Counters that trusted execution environments make phones hard to hack, then introduces the ancient hybrid idea, starting with the Greek ostracon that gave English the word ostracize.

The medieval split tally stick from hazelwood, notched then split into the longer stock and shorter foil so only the original halves fit back together.

The split tally was legal proof in medieval courts, referenced in the Napoleonic Code article 1333, and still used in rural 20th century Switzerland.

You can build a voting tally stick with blockchain and paper, where a phone vote in a trusted execution environment tied to a national DID produces both a blockchain record and a paper receipt.

The paper receipt is linked to the DID plus phone hardware so only matching pieces fit, and tampering with the blockchain would make the paper tallies disagree.

At audit you count paper ballots against blockchain ballots and they must match like a UTXO system's inputs and outputs, optionally with two physically linked receipts.

Voters could instead vote at a booth creating a blockchain record checkable on their phone, a hybrid digital system only as good as the strength of Bob's ID.

The distributed design keeps blockchain in the cloud and paper in different places with independent counters, allowing a double or even tri tally system for high priority elections.

Bob can query the blockchain to verify his vote counted, giving inclusive accountability by checking registered voters against votes cast.

AnonCreds let a voter prove how they voted without revealing their name, useful for party registration rules, and calls phone voting inevitable via TEEs like Samsung Knox.

The primitives exist and MPC can layer on if you distrust the phone, but the hard value is wiring them together, and no state is above reproach without modernizing its ID system.

People call a national ID the mark of the beast yet accept social security cards, passports, the NSA, and putting everything on Facebook and Google.

The identity vulnerability makes identity theft a 50 billion dollar a year business, so you build in layers starting with a secure self sovereign national ID on the DID standard.

After DIDs come secure registration and phone voting, then a low cost blockchain and paper ballot layer with single or dual tally stick receipts, and uncorroborated votes go uncounted at audit.

EV voting is one of the hardest cryptographic protocols, intrinsically adversarial and resource constrained, spanning confidentiality, integrity, non repudiation, and authentication.

Blockchain gives a brokering mechanism, a tamper resistant timestamped PKI, and a place to tether digital and paper receipts via secrets and nonces, buildable in one to three years by 5 to 10 cryptographers.

Reiterates IO will never bid on election work without also owning the identity infrastructure, or you just build the most secure counterfeit machine.

Uses the birther conspiracy to illustrate garbage in garbage out: before Medicare, immigrants could pay a doctor to file a legitimate birth certificate for a foreign born child.

A legitimate breeder document does not mean accurate data, since it presupposes the prior stage was correct, and voter registration is the case study.

You must solve identity, then registration, then elections in strict order, and urges skepticism toward anyone claiming blockchain alone solves voting.

Cardano solved the easier case with CIP 1694, where ADA holders are pre authenticated via a public private key system, the gold standard of identity like PGP.

Concludes physical elections still need paper and a tally stick concept, noting humanity has used tally sticks for over 3,000 years including in ancient China.