Pseudorandom Function (PRF): Difference between revisions

From Cryptology City
Jump to navigation Jump to search
No edit summary
No edit summary
Line 18: Line 18:
=== Security ===
=== Security ===
A PRF is ''secure'' if for all efficient <math>D</math>, there exists a negligible function <math>\nu</math>, such that
A PRF is ''secure'' if for all efficient <math>D</math>, there exists a negligible function <math>\nu</math>, such that
 
<center><math>
<math>
\big| \Pr [D^{F_k}(1^{\lambda}) = 1: k\gets \mathsf{Gen}(1^{\lambda})] -
\big| \Pr [D^{F_k}(1^{\lambda}) = 1: k\gets \mathsf{Gen}(1^{\lambda})] -
\Pr [D^{R}(1^{\lambda}) = 1 : R \gets \mathcal{F}[\mathcal{D},\mathcal{R}]] \big| \le \nu(\lambda).
\Pr [D^{R}(1^{\lambda}) = 1 : R \gets \mathcal{F}[\mathcal{D},\mathcal{R}]] \big| \le \nu(\lambda).
</math>
</math></center>


=== Weak Security ===
=== Weak Security ===
A PRF is ''weakly secure'' if for all efficient <math>D</math>, and all polynomials <math>s</math>, there exists a negligible function <math>\nu</math>, such that
A PRF is ''weakly secure'' if for all efficient <math>D</math>, and all polynomials <math>s</math>, there exists a negligible function <math>\nu</math>, such that
 
<center>
<math>
<math>
\big| \Pr_{x_i \gets \mathcal{D}} [D((x_i,F_k(x_i))_{i\in [s(\lambda)]}) = 1] -
\big| \Pr_{x_i \gets \mathcal{D}} [D((x_i,F_k(x_i))_{i\in [s(\lambda)]}) = 1] -
\Pr_{x_i \gets \mathcal{D}} [D((x_i,R(x_i))_{i\in [s(\lambda)]}) = 1] \big| \le \nu(\lambda).
\Pr_{x_i \gets \mathcal{D}} [D((x_i,R(x_i))_{i\in [s(\lambda)]}) = 1] \big| \le \nu(\lambda).
</math>
</math>
</center>


The difference in this definition is that we sample $s(\lambda)$ inputs ''at random'' instead of allowing a distinguisher to choose them. This is strictly weaker, as a distinguisher could always query their oracle at random locations in the stronger definition.
The difference in this definition is that we sample $s(\lambda)$ inputs ''at random'' instead of allowing a distinguisher to choose them. This is strictly weaker, as a distinguisher could always query their oracle at random locations in the stronger definition.

Revision as of 22:28, 3 July 2024


A Pseudorandom Function (PRF) is a primitive that allows someone to succinctly represent a function that is indistinguishable from a random function. A user of a PRF generates a key, which it can use to evaluate a function at many points. Any efficient adversary, who only sees these inputs and outputs of the keyed function, cannot distinguish them from a random function.

PRFs were originally defined by [GGM84] and are a basic building block for many more complex primitives such as Symmetric Encryption (SE).

Formal Definition

Syntax

A Pseudorandom Function (PRF) is a tuple of functions , with respect to a keyspace , domain , and range , such that:

  • , takes a security parameter, and outputs a key ,
  • , takes a key and input , and outputs an element .

Generally, we assume that is "large," in the sense that it grows exponentially with the security parameter. If instead bounded by some polynomial in the security parameter, then the primitive is a "small-domain" PRF.

Security

A PRF is secure if for all efficient , there exists a negligible function , such that

Weak Security

A PRF is weakly secure if for all efficient , and all polynomials , there exists a negligible function , such that

The difference in this definition is that we sample $s(\lambda)$ inputs at random instead of allowing a distinguisher to choose them. This is strictly weaker, as a distinguisher could always query their oracle at random locations in the stronger definition.

Relationship to other primitives

There are also many folklore results surrounding pseudorandom functions:

  • PRPs over are PRFs, due to the switching lemma [[[folklore]]]

Sufficient assumptions

Variations

Other Notes