6.5620 (6.875), Fall 2022: Template

6.5620 (6.
875), Fall 2022
Homework # 1 Due: 21 September 2022, 11:59:59pm ET
• Typsetting: You are encouraged to use LATEX to typeset your solutions. You can use
the following template.
• Submissions: Solutions should be submitted to Gradescope.
• Reference your sources: If you use material outside the class, please reference your
sources (including papers, websites, wikipedia).
• Acknowledge your collaborators: Collaboration is permitted and encouraged in

small groups of at most three. You must write up your solutions entirely on your own
and acknowledge your collaborators.
Problems:
1. (5 points) Working with negligible functions. Recall that a non-negative function

ν : N R
→ + is negligible if it decreases faster than the inverse of any polynomial
(otherwise, we say that ν is non-negligible). More precisely, for all c > 0, there exists a
constant N > 0 such that for all n ≥ N , ν(n) < n−c .
State whether each of the following functions is negligible or non-negligible, and prove
your assertion. For all of the problems below, we take the base of the logarithm to be 2.
(a) ν(n) = 1/2100 log n .

(b) ν(n) = 1/2100 log n·log log log n .
(c) ν(n) = p(n) · µ(n), where p(n) = O(nk ) for some constant k, and µ(n) is a
negligible function. Either prove that ν is always negligible, or come up with a
counter-example.
1
(d) ν(n) = (µ(n)) p(n) , where p(n) and µ(n) are as defined in (c). Either prove that ν
is always negligible, or come up with a counter-example.
log∗ n
(e) ν(n) = 1/22 , where log∗ n is the number of times the logarithm function must
be iteratively applied to n before the result is less than or equal to 1. More
concretely, (
∗ 0 if n ≤ 1
log n := ∗
1 + log (log n) if n > 1.
You may use the fact that log∗ n ≤ n without proof. (Hint: How does log∗ n
compare to log log n?)
1
2. (11 points) Statistical and computational indistinguishability. We think of
distributions
P PY on a (finite) set Ω as functions X, Y : Ω → [0, 1] such that for
X,
ω∈Ω X(w) = ω∈Ω Y (ω) = 1. The statistical distance (also known as variational or
L1 distance) between X and Y is defined as
1X
∆(X, Y ) := |X(ω) − Y (ω)| .
2 ω∈Ω
(a) (3 points) Show that the following is an equivalent definition:
∆(X, Y ) := sup |X(A) − Y (A)| ,

A⊆Ω
P
where X(A) is shorthand for ω∈A X(ω).
(b) (3 points) Let D0 and D1 be two distributions over Ω. Suppose that we play the
following game with an algorithm A. First, we pick at random a bit b ← {0, 1}
and then we pick x ← Db and we give x to A. Finally, A returns a bit. It wins if
the bit returned is equal to b. Show that the highest success probability in this
game is exactly 12 + 21 ∆(D0 , D1 ).
(c) (1 point) Give the definition of computational indistinguishability using similar
language as in the previous question. (This part should not take more than 3-5
sentences.)
(d) (4 points) For a probability distribution D over Ω and positive integer m, let
Dm denote the product distribution over Ωm , obtained by drawing a tuple of m
independent samples from D. Let X = { Xn }n and Y = { Yn }n be ensembles of
distributions that are efficiently sampleable (in PPT), and let m(n) = poly(n) be
some fixed polynomial. Prove that if X and Y are computationally indistinguish-
c m(n) c m(n)
able, or, in symbols, X ≈ Y, then { Xn } ≈ { Yn }. (Where do you use that
Xn , Yn are efficiently sampleable?) This shows that if two efficiently sampleable
distributions are computationally indistinguishable given one sample, then they
are also computationally indistinguishable given polynomially many samples.
3. (8 points) PRG or not? Let G : {0, 1}2n → {0, 1}ℓ be a pseudorandom generator,
where ℓ ≥ 2n + 1. In each of the following, say whether Gc is necessarily a pseudo-
random generator. If yes, give a proof. Otherwise, show a counterexample. Your
counterexamples must rely only on the existence of pseudorandom generators.
(a) (2 points) Consider G0 : {0, 1}2n → {0, 1}ℓ , where G0 (s) := G(s). Here, s is the
bit-wise complement of s.
(b) (3 points) Consider G1 : {0, 1}n → {0, 1}ℓ , where G1 (s) := G(0n ||s).
(c) (3 points) Consider G2 : {0, 1}2n → {0, 1}2ℓ , where G2 (s) := G(s)||G(s + 1
mod 22n ).
2
4. (3 points) A PRG from a PRF. Prove that, if F is a length preserving pseudorandom
def
function, then G(s) = Fs (⟨1⟩)∥Fs (⟨2⟩)∥ . . . ∥Fs (⟨ℓ⟩), where ⟨i⟩ is the n-bit binary
representation of i, is a pseudorandom generator with expands ℓ bits to ℓ · n bits.
5. (8 points) A Locking Scheme.

Veronica claims to Lucy, her classmate in 6.5620, that she can read minds, but Lucy
does not believe it one bit, and is willing to bet $100 that it’s all baloney. They decide
to play a game where Veronica can prove to Lucy that she is a bonafide mind-reader.
If Veronica is lying, she should not be able to win in the game; and if Veronica indeed
possesses this supernatural power, Lucy should grant victory (and $100) to Veronica at
the end of the game (no matter how much she hates to lose). They decide to play the
following game.
(a) Veronica (we’ll call her V ) sends Lucy a string v (of a certain length that they
decided on), chosen at random.
(b) Lucy (we’ll call her L) chooses a random bit σ ∈ {0, 1}, and Veronica has to guess
what this bit is. To do this, Lucy sends
ℓ ← L(σ, v; r)
to Veronica, where r is Lucy’s private random coins that only she knows.
(c) Now, Veronica reads Lucy’s mind and guesses what σ is.
(d) Finally, Lucy “unlocks” her bit σ by sending σ and r to Veronica. Veronica verifies
that ℓ is indeed L(σ, v; r).
Veronica wins if her guess of σ is correct.
(a) (1 point) Veronica should not gain any information about Lucy’s bit from viewing
the lock (i.e. after step 2 of the game). In other words, a malicious (but compu-
tationally bounded ) Veronica V ∗ should not be able to learn anything about the
honest L’s choice bit σ, no matter what initial message v ∗ she sent.
Using the notion of indistinguishability, give a formal definition of this concealing
property of L.
(b) (1 point) Lucy should not be able to unlock her bit both ways, otherwise she can
always get away with not paying Veronica $100 even if Veronica is a mind-reader
(do you see why?) To ensure this, the locking algorithm L has to be “unmodifiable”.
Give a formal definition of this property.
(c) (2 points) Let G be any length-tripling function, i.e., one for which |G(x)| = 3|x|
for every x ∈ {0, 1}∗ . Give an upper bound on the probability, over the choice of
a random 3n-bit string v, that there exist two inputs x1 , x2 ∈ {0, 1}n such that
G(x1 ) ⊕ G(x2 ) = v.
3
(d) (4 points) Let G be a length-tripling PRG (which we have seen can be obtained
from any PRG). Use G to construct a secure locking scheme (i.e. define the
algorithms V and L), and prove that it is both concealing and unmodifiable
according to your definitions.

6.5620 (6.875), Fall 2022: Template

Uploaded by

Copyright:

Available Formats

6.5620 (6.875), Fall 2022: Template

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

6.5620 (6.875), Fall 2022: Template

Uploaded by

Copyright:

Available Formats

6.5620 (6.

875), Fall 2022

Homework # 1 Due: 21 September 2022, 11:59:59pm ET

• Submissions: Solutions should be submitted to Gradescope.

• Acknowledge your collaborators: Collaboration is permitted and encouraged in

1. (5 points) Working with negligible functions. Recall that a non-negative function

(a) ν(n) = 1/2100 log n .

(a) (3 points) Show that the following is an equivalent definition:

∆(X, Y ) := sup |X(A) − Y (A)| ,

5. (8 points) A Locking Scheme.

Veronica wins if her guess of σ is correct.

You might also like