14
R yan Henr y I538/B609: Introduction to Cryptography Fall 2015 Lecture 5

Ryan Henry I 538 /B 609 : Introduction to Cryptography

Embed Size (px)

Citation preview

Ryan Henry

I538/B609:

Introduction to CryptographyFall 2015 Lecture 5

Ryan Henry

Last Thursday’s lecture:• Perfectly secret encryption

Today’s lecture:• Negligible functions and probabilities

• Computationally secret encryption

Ryan Henry

Thursday, September 10Assignment 1 is due on Tuesday, September 8

(that’s this Thursday!)

Assignment 2 has been posted

2

Ryan Henry

Defining computational secrecy

▪ Recall: Prefect secrecy reveals nothing about plaintext– “Bad News” lemma: perfect secret 1M1 = 1C1 = 1K1– “Bad News” corollary: can only encrypt one plaintext per key

▪ Compromise: In practice, it is sufficient for Enck(m) to reveal

“essentially nothing” about plaintext to “real world” attackers– Real world attackers == attackers with only bounded resources

1

Ryan Henry

Defining “real world attackers”

▪ Idea 1: can rents 1,000 Amazon EC2 instances for 100 years – Attacker’s PhD thesis proposes a faster algorithm

▪ Idea 2: Attacker spends 10 million USD on hardware– Intel releases a significantly faster CPU (or GPU)

▪ Idea 3: Attacker controls 1.5 million host botnet– Botnet grows to contain 150 million hosts

5

Ryan Henry

Defining “real world attackers”

▪ The “right” idea: Adversary is an arbitrary Turing Machine that runs in polynomial time– We make no assumption about which polynomial

– Prove that attacker’s success probability is insignificantly small

6

Defn (Concrete security): An encryption scheme (Gen, Enc, Dec) is said to be (t , ε)-secure if every Probabilistic Turing Machine that halts

after t steps can “break” the secrecy of (Gen, Enc, Dec) with probability at most ε.

Ryan Henry

Turing Machines

7

▪ A simple, well-defined mathematical model of computation

▪ Measure running time by number of steps a TM requires before it halts– Measure robust in that all other “reasonable” models

of computation require “polynomially related” number of steps

Church-Turing thesis: TMs are universal: anything you can compute in theory, you can compute on a TM!

Ryan Henry

Probabilistic polynomial time (PPT)

Defn: A TM runs in polynomial time (PPT) if, on input an n-bit string , it halts after (at most) O( t(n) ) steps, where t(∙)

denotes some polynomial function.

8

Defn: A polynomial time TM is said to be probabilistic

polynomial time (PPT) if its output is a random variable.

Ryan Henry

Efficient attackers / algorithms

▪ An attacker is said to be efficient if we can implement is using a PPT Turing machine

Q: Why equate “efficient” with “probabilistic polynomial time”?

A: Experience tells us “doable in polynomial time” roughly equivalent to “doable (eventually) in practice”

Nice composition theorems:– poly(n) + poly(n) = poly(n) ← deg(f(n) + g(n)) =– poly(n) * poly(n) = poly(n) ← deg(f(n) * g(n)) =– poly( poly(n) ) = poly(n) ← deg(f(g(n))) =

9

??

??

??

deg(f)+deg(g)

max{ deg(f), deg(g) }

deg(f)*deg(g)

Ryan Henry

Negligible functions

Defn: A function is negligible if for every c > 0.

10

▪ Equivalently:

– A function is negligible if it “vanishes” faster than the inverse of

every positive polynomial

– A function is negligible if such that for all

Ryan Henry

Noticeable functions

11

Q: If a function is not negligible, is it necessarily noticeable?

A: No! See question 2 on assignment 2!

Defn: A function is noticeable if for some c > 0.

Ryan Henry

Closure for negligible functions

▪ Fact 1: If and are negligible functions, then is a negligible

function.

▪ Fact 2: If is a negligible function and is a positive

polynomial, then is a negligible function

▪ Fact 3: If is a negligible function and if a constant, then is a

negligible function

12

Ryan Henry

Negligible and overwhelming probabilities

Defn: An event E occurs with probability negligible in n if Pr[E]is (bounded above by) a negligible function of n.

13

Defn: An event E occurs with probability overwhelming in n if Pr[E]is (bounded above by) a negligible function of n.

Ryan Henry

That’s all for today, folks!

14