Hide a secret in the constant term of a random polynomial; hand out points ON that curve as SHARES. Any k of them pin down the whole polynomial (and recover the secret) by interpolation — but k−1 shares reveal literally nothing, because infinitely many curves pass through too few points. It’s how you split a master key so 3 of 5 officers must agree, no single one can act alone. Slide to gather shares and watch the secret snap into focus.
Shamir’s (k,n) secret sharing hides a secret s as the constant term of a random degree-(k−1) polynomial over a prime field; each participant gets one point (x, f(x)) as a SHARE. Any k shares uniquely determine the polynomial (Lagrange interpolation at x=0 recovers s), while any k−1 shares leave every possible secret equally likely — information-theoretic security, not computational. It gives threshold trust: m-of-n control of a key, distributed backups, and is a building block of multiparty computation and threshold signatures. A fail-loud self-check throws unless 3 shares of a degree-2 sharing reconstruct the secret. ◆ real cryptography, node-verified.
Security is exact ONLY with a truly random polynomial and a proper field; it protects confidentiality, not integrity (a lying dealer/participant needs verifiable secret sharing). The k-of-n threshold and Lagrange recovery are exact.