Research
Timechain
Search Donate
Sponsors
Cake Wallet, private Bitcoin and Monero walletSideShift.ai, no-signup crypto exchangeOrangeRock, non-custodial bitcoin spot and perps tradingBraiins Hashpower, buy and sell Bitcoin hashrateLayer Two Labs, drivechains for Bitcoin
S10 · E7 · 2:14:12

S10 E7: Yuval Kogman (nothingmuch) on Improving Bitcoin Privacy

22 Feb 2022 · by Vlad Costea · 7 min read
Published at block 724,462
Cover art for S10 E7, Yuval Kogman (nothingmuch) on Improving Bitcoin Privacy
MP3 (184 MB)

The audio keeps playing while you browse the site.

Read the transcript16,446 words, every line links to the minute it was said

Ask this episode

Embed this episode

Paste this on your site or blog. It is a small player that streams from here: no trackers, no cookies, no ads.

Preview the player

When I visited the Wasabi Wallet HQ in Budapest in November 2021, the person I most wanted to keep talking to was a coder who goes by nothingmuch on Twitter. His name is Yuval Kogman, he has been writing code for more than two decades, and at the time he was helping with the WabiSabi CoinJoin protocol behind what was going to become Wasabi 2.0. Off the record he was chatty. Once I pressed the record button, he became "so bad at this stuff."

That shyness lasted about two minutes. What followed was one of the most rigorous conversations about Bitcoin privacy I have recorded: what privacy actually protects, why blockchain analysis works better than it should, how WabiSabi credentials work, and why Yuval said a firm "No" when I asked whether an emergency hard fork should also bring confidential transactions to Bitcoin.

Privacy From Whom?

I started with the usual framing about counterparties and third parties. Yuval accepted it as a starting point and then widened it. Bitcoin removes the intermediary who can authorize or refuse a payment, but to resist control and inflation the whole ledger has to be transparent, so "the whole world is observing." His conclusion: "you need privacy from the rest of the world," and also from whoever pays you and whoever you pay.

The reasons differ. A business needs confidentiality. Someone living "in a part of the world where the social contract is fragile" may put themselves at risk of theft just by revealing how much they hold. Most of the Bitcoin conversation focuses on KYC exchanges and dragnet surveillance, which mainly hurt fungibility. Yuval cares about that, but the personal safety side "personally concerns me to the same degree."

He also warned against treating any single tool as a solution. Dust attacks and the old rebroadcasting behavior in Bitcoin Core were his examples of small leaks with long consequences. "Just using Tor is not enough, right?" The way tools compose matters more than the tools themselves, and "you should always assume that the power of the adversary is strictly improving in time."

Satoshi Saw the Problem, but Not How Fragile It Was

Satoshi's whitepaper already describes the two heuristics chain analysis still relies on: assume all inputs of a transaction belong to the same owner, then guess which output is the change. Yuval called it "a very naive and crude way of trying to recover information, but nonetheless, we have very good reason to believe that it's still highly effective."

The result that impressed him most was Jonas Nick's master thesis, which exploited a Bloom filter privacy leak in light clients (since responsibly reported and mostly fixed) to measure how well those heuristics actually perform. The answer was a recall rate above 80%. "And this is kind of a terrifying result." The irony, as Yuval put it, is that at the consensus layer individual satoshis inside a transaction are perfectly fungible. The flaws "mainly have to do with how we apply the protocol."

Why Money Needs Privacy

I asked the question every privacy guest gets: why should anyone care? Yuval answered like an economist. Markets are how people with private knowledge coordinate without a central planner, and fiat has corrupted that signal so badly that the argument for destroying privacy "to protect the public just falls flat" for him. He went further back, citing the line Nick Szabo borrowed from Richard Dawkins in Shelling Out: money as a token of delayed reciprocal altruism.

If every counterparty can see your strategy, every trade becomes a game of exploiting it. "So I think that privacy is fundamentally important for the function of money."

Then I blew the dog whistle and brought up Bitcoin cash. Not BCH, but paper bills backed by BTC, the kind of physical money I think Bitcoin needs. Yuval had no problem with it. He mentioned bearer devices like Opendime and Tangem, and centralized eCash, which Hal Finney had already described in early posts on Bitcoin Talk. In his view, more choice is "strictly better." We already have a kind of eCash, he said: custodial exchanges, just "a very draconian, like, panopticon version of that." The danger is a monopoly: "if there is a single centralized Chaumian eCash server that just manages all of the Bitcoin, that would be a terrible, terrible outcome."

Envelopes Full of Sand

This was the segment I had invited him for. For CoinJoins, Yuval explained, "WabiSabi is merely a denial of service protection mechanism," but it achieves that through an eCash-like system of anonymous credentials, building on David Chaum's blind signatures and on the homomorphic amount commitments that confidential transactions also use.

His analogy deserves to be stolen. You walk into a room with envelopes filled with sand. The coordinator weighs the envelopes on its own scale, sees that the total is right, and seals them with wax without knowing how the sand is split between them. Later you come back, show valid seals, the weight checks out, and the sand is spilled into one pile so nobody can tell where each grain came from. Each credential is "a little bit like a UTXO with no transaction history." Users can register outputs only for the amounts they are entitled to, and the coordinator cannot link them back to the inputs.

The flexibility is also the risk. A protocol that is "literally just a PSBT whiteboard" can produce brittle privacy if used naively. "I mean, it's a giant foot gun, basically." His answer was a mix of standard denominations (powers of two, the 1-2-5 series familiar from fiat cash, and powers of three with their doubles) that can add up to almost any amount, so large transactions stay ambiguous without forcing everyone into one rigid output size.

The Stigma Problem

I relayed the usual criticism: CoinJoins don't scale, and anyone can see that one happened. Yuval separated the two. The coordination problem is solved well enough in practice, and a UTXO that keeps registering without signing can simply be banned. Blocks are the real constraint, which is why layer two matters so much.

The social problem worried him more. Today, opting into privacy is "a use it or lose it proposition," and doing it openly is a statement: "I deserve privacy and I'm claiming it." If CoinJoined coins and KYCed coins end up as two separate halves of the UTXO set that never interact, he called that "a terrible, terrible outcome." He also explained the personal risk of an exchange closing your account after you mix coins you withdrew from it. I agree with him that this is a political fight more than a technical one, and that the only way to lose it is to stop using the tools.

No to Confidential Transactions in an Emergency Hard Fork

A week before recording, I had asked Yuval this question off-air and he'd said the short answer was no. On tape, I put it to him properly. If Bitcoin ever needs an emergency hard fork, should it also add confidential transactions, ring signatures or Mimblewimble?

"No. There's several reasons for this." The first is that Monero and Zcash can't really be pruned, so the cost of validating them only grows. The second is inflation resistance, "a crucial property that to me is more important than privacy, because privacy can be fixed under this constraint." He reminded me that "both Monero and Zcash had subtle cryptographic bugs relating to inflation resistance." If we already got the simple version wrong, he argued, making the arithmetic far more delicate doesn't add up.

His alternative is layers. Lightning and payment channels already hide individual payments and reduce what reaches the chain. "The best privacy is having the least amount of information on the blockchain," so scaling and privacy point in the same direction. He summed it up with a line he had tweeted: "layer one Bitcoin is for enemies." And on the protocol itself: "And I think it's not broken enough, right? Like we can fix it."

I liked his pragmatism, and I told him so. Confidential transactions have a simple design and have been tested on Liquid, but someone who has studied every one of these mechanisms and still argues for caution deserves to be heard.

A Perl Past and Merkleized Obsessions

For the last few minutes, I asked where people can follow him. Twitter, which he called "a truly despicable, toxic hellhole of a software" before admitting that people there are on average cool and nice. His GitHub, he warned, is mostly dead code from his years in the Perl community. These days he says he is "obsessed, utterly obsessed, with content-addressed systems."

His reason for working on all this was the least technical thing he said: "I just want to think more nice thoughts before I die."

If you want more on how WabiSabi reached users, listen to my conversation with Lucas Ontivero in S10 E5, browse the CoinJoin and Bitcoin privacy topic pages, or read the full transcript of this episode, which is worth it for the sand analogy alone.

Listen to Yuval Kogman on Spotify, Apple Podcasts, and More

Time stamps:

  1. 00:00:29– Intro: meeting nothingmuch in Budapest
  2. 00:02:48– Privacy from counterparties and from the world
  3. 00:08:05– Dust attacks, rebroadcasting and Tor
  4. 00:14:30– Satoshi's heuristics and Jonas Nick's thesis
  5. 00:19:05– Why privacy matters for money
  6. 00:32:12– Bitcoin-backed cash and eCash
  7. 00:37:54– How WabiSabi works
  8. 00:51:05– The envelopes and sand analogy
  9. 00:57:16– Refinements, standard denominations and ambiguous CoinJoins
  10. 01:12:05– Does CoinJoin scale?
  11. 01:25:39– Stigma and a fragmented UTXO set
  12. 01:37:42– Would you add privacy in an emergency hard fork?
  13. 01:52:45– Layer one is for enemies
  14. 01:57:45– Where to follow Yuval

Verify this episode

This is the SHA-256 fingerprint of the MP3 as published here, timestamped in the Bitcoin blockchain with OpenTimestamps. If one byte of the file changes, the fingerprint changes.

SHA-256
d3c742fd8aa58998d4611c279550675de3827fad6100a14c13fb7a98a520f678
Proof
2022-02-15_s10_e7_yuval_kogman_on_improving_bitcoin_privacy_-_15_02_2022__12_01.mp3.ots · all episode hashes (proof)
Status
Anchored in Bitcoin block 974,312
Timestamped
After release. The episode came out on 15 Feb 2022, but this proof was made on 3 Nov 2026 and anchored in block 974,312, so it proves this exact file existed by 3 Nov 2026, not on the release date.
How to check it yourself
  1. Download the MP3 and the proof into the same folder.
  2. Run sha256sum 2022-02-15_s10_e7_yuval_kogman_on_improving_bitcoin_privacy_-_15_02_2022__12_01.mp3 (on a Mac: shasum -a 256 2022-02-15_s10_e7_yuval_kogman_on_improving_bitcoin_privacy_-_15_02_2022__12_01.mp3). The result must match the SHA-256 above.
  3. Install the OpenTimestamps client (pip install opentimestamps-client) and run ots verify 2022-02-15_s10_e7_yuval_kogman_on_improving_bitcoin_privacy_-_15_02_2022__12_01.mp3.ots. It hashes the MP3 again and checks the proof against your own Bitcoin node, then prints the block and the date the file existed by. Without a node, ots info 2022-02-15_s10_e7_yuval_kogman_on_improving_bitcoin_privacy_-_15_02_2022__12_01.mp3.ots shows the block height to look up in any block explorer.

Share: X · Email · Link

Related

Podcast Season Preview Bitcoin artwork

Preview: BTCTKVR Podcast Season 10

26 Jan 2022 · 10 min read

The Bitcoin Takeover Podcast keeps on keeping it real with season 10. This time around, we have 16 guests whose topics of…

Comments

Comments run on my own server. No trackers, no tracking cookies.

0:000:00