program derivation of operations in finite fields of prime order
DESCRIPTION
The higher-quality version of my undergraduate research presentation (still some typos and missing formulas, but better explanations, more pleasing layout, more coherent analysis, etc.). Presented at Oklahoma Computing Consortium 2011.TRANSCRIPT
Introduction Title
Program Derivation of Operations in Fp
Charles Southerland Dr. Anita Walker
Department of Mathematics & Computer ScienceEast Central University
Oklahoma Computing Consortium Conference 2011
Southerland, Walker Program Derivation of Operations in Fp
Introduction Thanks
Special Thanks
I would like to say a special thank you to:
Dr. Anita Walker for working closely with me throughoutthis project, and for introducing me to abstract algebra
Dr. Bill Walker for introducing me to program derivation
Prof. Clay Carley for working with me on cryptology, whichfirst lead me to this particular problem
The creators of Beamer for allowing LATEX to save me fromthe abyss of WYSIWYG presentation software
Southerland, Walker Program Derivation of Operations in Fp
Finite Fields Outline
Outline
1 Finite FieldsDefinitionField OrderA Well-Known Finite Field
2 Program Derivation
3 Multiplicative Inverse in Fp
Southerland, Walker Program Derivation of Operations in Fp
Finite Fields Definition
The Definition of a Field
Definition
A field is a 3-tuple of a set F and two operations (called additionand multiplication) for which certain properties hold:
Closure of F under both operations
Associativity of both operations
Distinct identities in F for the operations
Additive inverses for all items in F
Multiplicative inverses for all but the additive identity
Commutativity of both operations
Distributivity of multiplication over addition
Southerland, Walker Program Derivation of Operations in Fp
Finite Fields Definition
The Galois Field
A finite field is a field in which the contained set has finitecardinality (e.g., the field has a finite order).
All finite fields of the same order are isomorphic (so they are,for all practical purposes, the same).
Another name for a finite field is a Galois field.
Generalized fields are often denoted as F, but finite fields inparticular are usually denoted either with GF , GF (q), or Fq,where q is the order of the field.
Southerland, Walker Program Derivation of Operations in Fp
Finite Fields Field Order
The Order of a Finite Field
There exists a finite field of order q iff q = pn, where p isprime and n ∈ N..
When n = 1, Fp is isomorphic to (Zp,⊕,⊗) (the integersmodulo p with modular addition and modular multiplication).
When n > 1, Fpn is isomorphic to the splitting field off (x) = xpn − x over Fp.
This project focuses on fields of prime order, so I’m afraidthere will be no more discussion of Fpn .
Southerland, Walker Program Derivation of Operations in Fp
Finite Fields A Well-Known Finite Field
A Well-Known Finite Field of Prime Order: F2
Since 2 is prime, there is a finite field F2, and it has the form(Z2,⊕,⊗).
The operations are defined as:
Addition
⊕ 0 1
0 0 11 1 0
Multiplication
⊗ 0 1
0 0 01 0 1
As you can see, F2 is binary with XOR as addition and ANDas multiplication.
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Outline
Outline
1 Finite Fields
2 Program DerivationHistoryDijkstra’s Guarded Command LanguageWeakest Precondition Predicate TransformerThe Program Derivation Process
3 Multiplicative Inverse in Fp
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation History
The History of Program Derivation
Hoare’s 1969 paper An Axiomatic Basis for ComputerProgramming effectively launched the Formal Methodssubfield of CS.
Dijkstra’s paper Guarded Commands, Nondeterminacy andFormal Derivation of Programs introduced many of the ideaspresented in this paper.
Gries’ book The Science of Programming brings Dijkstra’spaper to a level undergrad CS and Math majors canunderstand.
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Dijkstra’s Guarded Command Language
Some Familiar Parts of Dijkstra’s Language
Variable Assignment
x := 1
Addition
x := x + y
Command Concatenation
b := b − a; x := x + y
Procedure Call
c := gcd(a, b)
Subtraction
b := b − a
Skip, then Abort
skip; abort
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Dijkstra’s Guarded Command Language
Dijkstra’s Guarded Commands
Guarded if-Block
if a > 0 → c := 2ut b > 0 → c := 3; a := 5ut c > 0 → c := 1ut c = 6 → c := 4fi
Guarded do-Block
do b = 0 → c := 1ut a > 0 → a := a− 1ut b < 4 → b := b + 1ut c = 1 → a := a− 1
od
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Dijkstra’s Guarded Command Language
A Famous Example
Greatest Common Divisor
proc gcd(a, b) ≡do a > b → a := a− but b > a → b := b − a
odreturn a.
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
The Weakest Precondition Predicate Transformer
Definition
The Weakest Precondition Predicate Transformer (wp) isdefined as follows:wp : P × L → L
P is the set of all finite-length programs
L is the set of all statements about the state of a computer
wp(s, r) = q
q is the weakest precondition (the initial state)
s is the program to be executed (which changes the state)
r is the postcondition (the resulting state)
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
wp and Dijkstra’s Language
Skip
wp(”skip”, r) = r
Command Concatenation
wp(”b := a; x := y”, r)= wp(”b := a”,wp(”x := y”, r))
Abort
wp(”abort”, r) = F
Variable Assignment
wp(”x := y”, r)= defined(y) ∧ r xy
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
wp and Dijkstra’s if-Block
Dijkstra’s if-Block
wp(”if a > 0 → c := 2ut b > 0 → c := 3; a := 5ut c > 0 → c := 1ut c = 6 → c := 4 fi”, r)
= (a > 0 ∨ b > 0 ∨ c > 0 ∨ c = 6)∧(a > 0 =⇒ wp(”c := 2”, r))∧(b > 0 =⇒ wp(”c := 3; a := 5”, r))∧(c > 0 =⇒ wp(”c := 1”, r))∧(c = 6 =⇒ wp(”c := 4”, r))
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
wp and Dijkstra’s do-Block, Part I
Let’s call this ”DO”:
do b = 0 → c := 1ut a > 0 → a := a− 1ut b < 4 → b := b + 1ut c = 1 → a := a− 1
od
Also, let’s call this ”IF”:
do b = 0 → c := 1ut a > 0 → a := a− 1ut b < 4 → b := b + 1ut c = 1 → a := a− 1
od
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
wp and Dijkstra’s do-Block, Part II
We define Hn(r) for n ∈ N and r ∈ L as:
For n = 1
H1(r) = (b 6= 0 ∧ a ≤ 0 ∧ b ≥ 4 ∧ c 6= 1) ∧ r
For n > 1
Hn(r) = H1(r) ∨ wp(”IF ”,Hn−1(r))
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation Weakest Precondition Predicate Transformer
wp and Dijkstra’s do-Block, Part III
Dijkstra’s Guarded do-Block
wp(”do b = 0 → c := 1ut a > 0 → a := a− 1ut b < 4 → b := b + 1ut c = 1 → a := a− 1 od”, r)
= (∃n ∈ N)Hn(r)
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation The Program Derivation Process
Program Derivation
Program Derivation
Given a precondition q ∈ L and a postcondition r ∈ L,derive a program s ∈ P that satisfies q = wp(s, r).
Southerland, Walker Program Derivation of Operations in Fp
Program Derivation The Program Derivation Process
Program Derivation Tips
Gather as much information as possible about theprecondition and postcondition.
Reduce the problem to previously solved ones wheneverpossible.
Look for a loop invariant that gives clues on how toimplement the program.
If you are stuck, consider alternative representations of thedata.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Outline
Outline
1 Finite Fields
2 Program Derivation
3 Multiplicative Inverse in Fp
Multiplicative InversesThe Greatest Common DivisorExploring Bezout’s IdentityProgram to Find the Multiplicative Inverse in Fp
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Multiplicative Inverses
Multiplicative Inverses in Fields of Infinite and Finite Order
Finding multiplicative inverses in a field of infinite order istypically not a problem.
Example
In (Q,+,×), multiplicative inverses are reciprocals (e.g., a−1 = 1a ).
Finding multiplicative inverses in fields of finite order can gettricky.
Example
In (Zp,⊕,⊗), multiplicative inverses are found using Bezout’sIdentity (i.e., ax + py = 1), which has two unknown values.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Multiplicative Inverses
Obtaining the Multiplicative Inverse from Bezout’s Identity
Noting that a and b are coprime (since b = p, and p is prime),gcd(a, b) = 1. So:
ax + by = gcd(a, b)ax + by = 1ax = by + 1ax = py + 1ax = 1
By the definition of multiplicative inverses, x = a−1.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp The Greatest Common Divisor
The Greatest Common Divisor
Recall the greatest common divisor program:
Greatest Common Divisor
proc gcd(a, b) ≡do a > b → a := a− but b > a → b := b − a
odreturn a.
This implementation was discovered by exploring the property:gcd(a, b) = gcd(a− b, b) = gcd(a, b − a)
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp The Greatest Common Divisor
The Loop Invariant of gcd
The loop invariant used in the primary loop of this program isgcd(a, b) = gcd(A,B).
The loop will exit when a = b, which occurs whena = b = gcd(a, b).
Since every iteration decreases the value of either a or b, theloop will progress toward termination (the loop is bound by(a− gcd(a, b)) + (b − gcd(a, b))).
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Exploring Bezout’s Identity
Bezout’s Identity and the gcd Property
Combining Bezout’s Identity with the gcd property, we get:
ax + by = gcd(a, b)= gcd(a, b − a)= au + (b − a)v= au + bv − av= a(u − v) + bv
So x ≡ u − v (mod b) and y ≡ v (mod a).As gcd is commutative, we derive a corresponding result if weexplored gcd(a− b, b) instead of gcd(a, b − a).
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Exploring Bezout’s Identity
Reassigning x and y as Linear Combinations: Part I
Each time the arguments of gcd get closer to their final value, it isshown that x is equivalent (mod b) and y is equivalent (mod a)to a linear combination of their corresponding values from Bezout’sIdentity after a and b have been modified as described in the gcdprogram.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Exploring Bezout’s Identity
Reassigning x and y as Linear Combinations: Part II
Specifically, it can be seen that x always has a positive coefficientof following corresponding values of x and a negative coefficient ofcorresponding values of y . Likewise, y always has a negativecoefficient of corresponding values of x and a positive coefficient ofcorresponding values of y .
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Exploring Bezout’s Identity
Reassigning x and y as Linear Combinations: Part III
Once the arguments to gcd are equal to each other (and equal tothe result of gcd), we can find the original values of x and y bymultiplying the coefficients that have been stored by the finalcorresponding values of x and y . However, since we are looking fora multiplicative inverse in Fp, we know gcd(a, p) = 1 as p is prime.Since this will give us x = 1 by simplification after using the gcdproperty one last time, we see that the y components areinconsequential.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Exploring Bezout’s Identity
Reassigning x and y as Linear Combinations: Part IV
Finally, we see that only the x coefficients are of any consequenceto the final result. Specifically, once the gcd algorithm is complete,since the initial (and desired) value of x can be found bymultiplying the final corresponding value of x by the propercoefficient of x , and since the final corresponding value of x = 1,we get that the desired value of x is equal to the coefficient of thecorresponding final value of x .
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Program to Find the Multiplicative Inverse in Fp
Finding the Loop Invariant
Based on the long-winded previous slides, we can describe a loopinvariant:
Axl + Byl = gcd(a, b)
where xl is is the linear combination that the initial value of x isequal to, and yl is the linear combination that the initial value of y .This loop invariant is nice, as it is fully compatible with the loopinvariant of gcd, and so it also progresses toward termination andhas a bound function that differs from that of gcd linearly.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Program to Find the Multiplicative Inverse in Fp
A Last Look at gcd for Reference...
Greatest Common Divisor
proc gcd(a, b) ≡do a > b → a := a− but b > a → b := b − a
odreturn a.
Southerland, Walker Program Derivation of Operations in Fp
Multiplicative Inverse in Fp Program to Find the Multiplicative Inverse in Fp
Multiplicative Inverse Program
Multiplicative Inverse
proc multinv(a, b) ≡xx := 1; yx := 0do a > b → a := a− b; yx := yx + xxut b > a → b := b − a; xx := xx + yx
odreturn xx .
Southerland, Walker Program Derivation of Operations in Fp
Conclusion Summary
Summary
Finite fields are very useful mathematical constructs that canbehave very differently from fields of infinite order.
Program derivation is performed by using the rules of theweakest precondition predicate transformer to determine whatsequence of conditions (and thus what program statements)must have occured between a given precondition andpostcondition.
While the process of deriving my multiplicative inverseprogram was time-consuming and complicated, the resultswere well worth the effort.
Southerland, Walker Program Derivation of Operations in Fp
Conclusion Future Work
Future Work
Program Derivation of Exponentiation in Fp
Extend scope to include Fpn
Explore factorization techniques
Finish library and create graphical front end
Southerland, Walker Program Derivation of Operations in Fp
Conclusion Contact Me
Contact Information
You can email me at [email protected] if you have anyfurther questions or comments.
Southerland, Walker Program Derivation of Operations in Fp