Questions: Learning with Errors (LWE)

5 questions to test your understanding

Score: 0 / 5
Question 1 Short Answer

LWE samples look like (a, <a,s> + e mod q) where e is small noise. Without the noise, the problem is trivially solvable. Why does adding small noise make it hard?

Think about your answer, then reveal below.
Question 2 Multiple Choice

Regev's original LWE hardness reduction is quantum — it uses a quantum algorithm to connect LWE to worst-case lattice problems. Does this mean LWE is only hard for classical adversaries?

AYes — quantum adversaries can solve LWE efficiently
BNo — the quantum reduction shows that breaking LWE is at least as hard as solving worst-case lattice problems, even using a quantum computer. The reduction itself uses quantum techniques, but the conclusion is that LWE is hard for both classical and quantum adversaries (assuming worst-case lattice problems are hard for quantum computers, which is widely believed)
CThe quantum reduction has been replaced by a classical reduction, so the question is moot
DLWE security against quantum adversaries requires different parameters
Question 3 Multiple Choice

Ring-LWE replaces the random matrix A with a structured matrix derived from polynomial multiplication in Z_q[x]/(x^n + 1). What are the benefits and risks of this structure?

ARing-LWE is strictly more secure because polynomial multiplication is harder to invert
BBenefits: keys are n elements instead of n^2 (smaller), operations are O(n log n) via NTT instead of O(n^2) (faster). Risks: the algebraic structure might enable attacks not possible on unstructured LWE — the ring could have exploitable properties. Module-LWE (used in Kyber/ML-KEM) balances by using small matrices over the ring, getting most efficiency benefits while reducing structural risk
CRing-LWE eliminates the noise requirement, simplifying the scheme
DRing-LWE is identical to LWE but uses different notation
Question 4 True / False

In Regev's LWE-based encryption, the ciphertext is roughly twice the size of the plaintext, and decryption works by computing an inner product that cancels the error 'almost' perfectly, with a rounding step recovering the exact plaintext bit.

TTrue
FFalse
Question 5 Short Answer

The decisional LWE problem (distinguishing LWE samples from uniform random) is at least as hard as the search LWE problem (finding the secret s). Why is this relationship unusual compared to other cryptographic problems?

Think about your answer, then reveal below.