Questions: Homomorphic Encryption

5 questions to test your understanding

Score: 0 / 5
Question 1 Short Answer

In lattice-based HE, each ciphertext carries a 'noise' term that grows with each operation. What happens if the noise exceeds a threshold, and how does bootstrapping solve this?

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

RSA is multiplicatively homomorphic: E(m1) * E(m2) = E(m1 * m2). Why doesn't this make RSA a useful homomorphic encryption scheme?

ARSA multiplication is too slow for practical use
BRSA supports only multiplication, not addition. Without both operations, you cannot compute arbitrary functions — you're limited to multiplication chains. Fully homomorphic encryption requires both addition and multiplication (since any Boolean circuit can be built from AND and XOR/OR). Additionally, textbook RSA lacks semantic security, making even its multiplicative homomorphism a vulnerability rather than a feature
CRSA's homomorphic property only works for prime plaintexts
DThe homomorphic property disappears when RSA uses padding
Question 3 Multiple Choice

CKKS is an HE scheme designed for approximate arithmetic on real numbers. Unlike BFV/BGV (which compute on integers exactly), CKKS treats the noise as part of the computation, allowing some precision loss. Why is this useful?

ACKKS is faster because it uses smaller parameters
BMachine learning and statistical computations inherently involve floating-point approximations. CKKS encodes real numbers and performs additions and multiplications that preserve values up to a controllable precision, matching the natural error tolerance of these applications. Exact schemes waste resources maintaining precision that the application doesn't need
CCKKS provides stronger security guarantees than exact schemes
DCKKS supports division while exact schemes do not
Question 4 True / False

FHE allows a cloud server to compute on encrypted data without learning anything about the data or the result. The client sends encrypted inputs and receives an encrypted result.

TTrue
FFalse
Question 5 Short Answer

Current FHE schemes are roughly 10,000-1,000,000x slower than computing on plaintext. What is the main source of this overhead?

Think about your answer, then reveal below.