Questions: Secure Multi-Party Computation

5 questions to test your understanding

Score: 0 / 5
Question 1 Short Answer

Three hospitals want to compute the average patient outcome across their databases without sharing individual patient data. Describe how MPC solves this and what security guarantee it provides.

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

The ideal/real paradigm defines MPC security by comparison to an ideal world with a trusted third party. Why is this approach better than listing specific attacks the protocol must resist?

AListing attacks is equivalent but less elegant
BThe ideal/real paradigm provides a universal security guarantee: anything that could go wrong in the real protocol could also go wrong in the ideal world (where the only possible 'attack' is choosing a bad input or learning from the output). This automatically protects against all possible attacks — known and unknown — without needing to enumerate them. A specific attack list would inevitably miss some attacks
CThe ideal/real paradigm is easier to prove but provides weaker guarantees
DThe paradigm only applies to protocols with three or more parties
Question 3 True / False

With an honest majority of participants (more than half are non-corrupted), MPC can achieve information-theoretic security without any computational assumptions.

TTrue
FFalse
Question 4 Multiple Choice

MPC for a function f guarantees that corrupted parties learn nothing beyond f's output. But what if f's output itself reveals sensitive information — for example, computing 'does any party's salary exceed $1M' reveals something about high earners?

AMPC prevents this by encrypting the output
BMPC cannot solve this — it guarantees the protocol leaks nothing beyond the output, but the function choice determines what the output reveals. If the function itself leaks too much, the parties must choose a different function. This is a function design problem, not a protocol problem
CAdding more rounds of interaction prevents output-based leakage
DDifferential privacy inside MPC solves this automatically
Question 5 Short Answer

Yao's garbled circuits enable two-party computation where one party 'garbles' the circuit and the other evaluates it. Why can't the evaluator learn intermediate wire values?

Think about your answer, then reveal below.