Lab: free tools · size first · stock scanner · crypto scanner · journal

Wallet Security

Shamir's Secret Sharing in Crypto

Shamir's Secret Sharing splits a secret into n shares so any t shares reconstruct it, while t-1 reveal nothing in theory. In crypto it is used for seeds and keys. Reconstruction puts the secret together in one place unless extra protocol prevents that.

Shamir's Secret Sharing splits a secret, such as a seed, into several shares so that any chosen number of them can rebuild it, while fewer reveal nothing. Some wallets build it into their backup options.

Split Is Not Signed; Reconstructed Is Hot

Shamir's scheme (1979) makes n shares of a secret with threshold t. Fewer than t shares are useless (in the information-theoretic story). t shares rebuild the secret, typically the seed. That rebuild is the moment the secret exists in RAM, on a laptop, or on a device. TSS tries not to have that moment. Shamir backups are about disaster recovery, not about daily signing, unless you like reconstructing on Tuesdays.

SLIP-39 is a common Shamir-like seed format in some hardware wallets. Social recovery apps may use Shamir or something that only looks like it. Ask whether reconstruction happens, where, and who can coerce t friends. Education only. Not a recommendation to split seeds.

Shamir split (schematic) Secret n shares t reconstruct

1. History that still binds Shamir

Before hardware wallets, people split paper seeds with XOR or envelopes. Shamir is the grown-up math for that instinct. Hardware vendors added SLIP-39 so you could put 2-of-3 plates in three cities. People then stored two shares in the same drawer. Math cannot fix a drawer.

TSS and Shamir got mashed in marketing because both say t-of-n. One reconstructs. One signs. Keep the verbs. History of losses includes both 'I lost too many shares' and 'I reconstructed on a malware PC.' For the asset-layer context, see Bitcoin.

2. Polynomials, formats, and operational rules

A polynomial of degree t-1 encodes the secret as the constant term; shares are points. t points interpolate. Implementation bugs, bad randomness, and leaking share photos are how theory meets GitHub. Use a well-reviewed implementation, preferably on an air-gapped device.

Do not photograph shares. Do not put t shares in one cloud. Do not test reconstruction on the daily driver if that machine is the threat. Write a map of where shares live and who can get them without you (inheritance) and who should not (everyone else).

3. How operators actually use Shamir

Honest jobs: 2-of-3 or 3-of-5 seed backup across locations for a cold wallet you rarely reconstruct; inheritance with a lawyer holding one share and not the map to the others without a trigger. Dishonest jobs: 3-of-5 with three shares on the same NAS; reconstructing every week to 'check.' Size the idea with the DennTech blog the same way you would any other crypto ticket: dollars of account risk first, notional second, leverage last.

Illustration only: 3-of-5 plates in five cities. Fire in one city is fine. Two nosy relatives in two cities plus a photo of a third in your email is not 3-of-5. Count independent failures, not plates. The self-custody is for unusual prints and price action, not for discovering that Shamir's Secret Sharing in Crypto exists.

Shamir event boxes Reconstruct on malware Shares too correlated

4. Failure modes

Correlated storage, lost shares below t, reconstruction leaks, fake SLIP-39 tools, and social-recovery guardians who get phished as a set. Also: mixing Shamir with a passphrase you then forget, which is two locks and no door. Related structure: Gemini.

5. Mistakes, limits, takeaways

Mistakes: Shamir = TSS; testing on a dirty PC; correlated shares. Limits: formats differ. Education only.

Not a recommendation to use Shamir. If you use it, drill reconstruction on a safe machine once, then stop.

The short version

  • Shamir splits a secret; t shares rebuild it in one place.
  • TSS signs without a rebuild. That's a different verb.
  • Correlated shares are one share.
  • Drill once on a clean machine.

The honest one-sentence object is: t-of-n share backups that reconstruct a seed or key. Information-theoretic security of t-1 shares assumes proper implementation. SLIP-39 is not BIP-39; do not mix wordlists casually. Educational only.

It is for cold-storage operators who will map independent failure domains. It is not for people who want t-of-n stickers without a reconstruction policy. Reconstructing weekly is a hot-wallet with extra steps.

Kraken/Gemini custody is a different object. Shamir splits are a self-custody tool. Photos of plates in iCloud are not 'offline.' That failure is usually faster than a support ticket and slower than a tweet.

Chain and venue context for Shamir's Secret Sharing in Crypto: off-chain backup math; on-chain still sees a normal key after reconstruct. Steel plates still fail to fire if they are in the same house.

It only changes how fast reconstruction on a compromised machine, or too many shares in one failure domain can arrive. Threshold too high (5-of-5) is how families lose coins.

Threshold too low (2-of-3 in one metro) is how burglars gain coins.

TSS can wrap Shamir for recovery of a share, and then you have two protocols to understand.

Passphrases plus Shamir is complexity; write it or do not use it. Coupons expire.

Inheritance drills should happen while you are alive.

Weekend 'help me recover' is a classic social-engineering hour.