But in cryptography, especially in , we rarely use pure Gram–Schmidt. Instead, we use its modified version (MGS) for numerical stability, or more importantly, the Gram–Schmidt coefficients : [ \mu_i,j = \frac\langle \mathbfv_i, \mathbfv_j^* \rangle\langle \mathbfv_j^ , \mathbfv_j^ \rangle ] These coefficients measure how much a basis vector leans on previous ones. Why Does Crypto Care? Lattice-based cryptosystems (like Kyber , Dilithium , or NTRU ) rely on the hardness of problems like SVP (Shortest Vector Problem) or CVP (Closest Vector Problem). To break them, an attacker tries to find a “good” basis — short and nearly orthogonal. That’s where the Lenstra–Lenstra–Lovász (LLL) algorithm enters.
Here’s a short on the Gram–Schmidt process in the context of cryptography and lattice reduction , as often referenced in challenges from CryptoHack (e.g., the “Gram Schmidt” module). Gram–Schmidt, Lattice Reduction, and Cryptography At first glance, the Gram–Schmidt process appears to be a purely linear algebra tool: given a set of vectors, it produces an orthogonal (or orthonormal) basis for the same subspace. However, in the world of post-quantum cryptography and lattice-based cryptanalysis , Gram–Schmidt plays a surprisingly central role. The Core Idea The classical Gram–Schmidt algorithm takes a basis ( \mathbfv_1, \dots, \mathbfv_n ) and outputs orthogonal vectors ( \mathbfv_1^ , \dots, \mathbfv_n^ ), where each ( \mathbfv_i^* ) is the component of ( \mathbfv_i ) orthogonal to the span of previous vectors. In exact arithmetic, this is deterministic and useful for computing volumes or QR decompositions. gram schmidt cryptohack