KGC_TEST/miracl/source/curve/pairing/zzn4.cpp

483 lines
9.9 KiB
C++

/***************************************************************************
*
Copyright 2013 CertiVox UK Ltd. *
*
This file is part of CertiVox MIRACL Crypto SDK. *
*
The CertiVox MIRACL Crypto SDK provides developers with an *
extensive and efficient set of cryptographic functions. *
For further information about its features and functionalities please *
refer to http://www.certivox.com *
*
* The CertiVox MIRACL Crypto SDK is free software: you can *
redistribute it and/or modify it under the terms of the *
GNU Affero General Public License as published by the *
Free Software Foundation, either version 3 of the License, *
or (at your option) any later version. *
*
* The CertiVox MIRACL Crypto SDK is distributed in the hope *
that it will be useful, but WITHOUT ANY WARRANTY; without even the *
implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. *
See the GNU Affero General Public License for more details. *
*
* You should have received a copy of the GNU Affero General Public *
License along with CertiVox MIRACL Crypto SDK. *
If not, see <http://www.gnu.org/licenses/>. *
*
You can be released from the requirements of the license by purchasing *
a commercial license. Buying such a license is mandatory as soon as you *
develop commercial activities involving the CertiVox MIRACL Crypto SDK *
without disclosing the source code of your own applications, or shipping *
the CertiVox MIRACL Crypto SDK with a closed source product. *
*
***************************************************************************/
/*
* MIRACL C++ Implementation file zzn4.cpp
*
* AUTHOR : M. Scott
*
* PURPOSE : Implementation of class ZZn4 (Arithmetic over n^4)
*
* WARNING: This class has been cobbled together for a specific use with
* the MIRACL library. It is not complete, and may not work in other
* applications
*
* Note: This code assumes that -2 is a Quadratic Non-Residue,
* so modulus p=5 mod 8
* OR p=3 mod 8
* OR p=7 mod 8, p=2,3 mod 5
*
* For example for p=3 mod 8 the representation is
*
* A+IB, where A=(a+ib), B=(c+id), I=sqrt(1+i)
* where i=sqrt(-1)
*/
#include "zzn4.h"
using namespace std;
ZZn4& ZZn4::powq(const ZZn2 &m)
{
zzn4_powq(m.getzzn2(),&fn);
return *this;
}
void ZZn4::get(ZZn2& x,ZZn2& y) const
{zzn2_copy((zzn2 *)&fn.a,x.getzzn2()); zzn2_copy((zzn2 *)&fn.b,y.getzzn2());}
void ZZn4::get(ZZn2& x) const
{zzn2_copy((zzn2 *)&fn.a,x.getzzn2()); }
void ZZn4::geth(ZZn2& x) const
{zzn2_copy((zzn2 *)&fn.b,x.getzzn2()); }
ZZn4& ZZn4::operator/=(const ZZn2& x)
{
*this*=inverse(x);
fn.unitary=FALSE;
return *this;
}
ZZn4& ZZn4::operator/=(const ZZn4& x)
{
*this*=inverse(x);
if (!x.fn.unitary) fn.unitary=FALSE;
return *this;
}
ZZn4& ZZn4::operator/=(int i)
{
if (i==2)
{
zzn2_div2(&fn.a);
zzn2_div2(&fn.b);
}
else
{
ZZn t=(ZZn)1/i;
zzn2_smul((zzn2 *)&fn.a,t.getzzn(),(zzn2 *)&fn.a);
zzn2_smul((zzn2 *)&fn.b,t.getzzn(),(zzn2 *)&fn.b);
}
fn.unitary=FALSE;
return *this;
}
ZZn4 inverse(const ZZn4& w)
{
ZZn4 y=w;
zzn4_inv((zzn4 *)&y.fn);
return y;
}
ZZn4 operator+(const ZZn4& x,const ZZn4& y)
{ZZn4 w=x; w+=y; return w; }
ZZn4 operator+(const ZZn4& x,const ZZn2& y)
{ZZn4 w=x; w+=y; return w; }
ZZn4 operator+(const ZZn4& x,const ZZn& y)
{ZZn4 w=x; w+=y; return w; }
ZZn4 operator-(const ZZn4& x,const ZZn4& y)
{ZZn4 w=x; w-=y; return w; }
ZZn4 operator-(const ZZn4& x,const ZZn2& y)
{ZZn4 w=x; w-=y; return w; }
ZZn4 operator-(const ZZn4& x,const ZZn& y)
{ZZn4 w=x; w-=y; return w; }
ZZn4 operator-(const ZZn4& x)
{ZZn4 w; zzn4_negate((zzn4 *)&x.fn,&w.fn); return w; }
ZZn4 operator*(const ZZn4& x,const ZZn4& y)
{
ZZn4 w=x;
if (&x==&y) w*=w;
else w*=y;
return w;
}
ZZn4 operator*(const ZZn4& x,const ZZn2& y)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator*(const ZZn2& y,const ZZn4& x)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator*(const ZZn4& x,int y)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator*(int y,const ZZn4& x)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator*(const ZZn4& x,const ZZn& y)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator*(const ZZn& y,const ZZn4& x)
{ZZn4 w=x; w*=y; return w;}
ZZn4 operator/(const ZZn4& x,const ZZn4& y)
{ZZn4 w=x; w/=y; return w;}
ZZn4 operator/(const ZZn4& x,const ZZn2& y)
{ZZn4 w=x; w/=y; return w;}
ZZn4 operator/(const ZZn4& x,int i)
{ZZn4 w=x; w/=i; return w;}
#ifndef MR_NO_RAND
ZZn4 randn4(void)
{ZZn4 w; zzn4_from_zzn2s(randn2().getzzn2(),randn2().getzzn2(),&w.fn); return w;}
#endif
ZZn4 rhs(const ZZn4& x)
{
ZZn4 w,A,B;
miracl *mip=get_mip();
int twist=mip->TWIST;
w=x*x*x;
A=(ZZn4)getA(); B=(ZZn4)getB();
if (twist)
{
if (twist==MR_QUARTIC_M)
{
w+=tx(A)*x;
}
if (twist==MR_QUARTIC_D)
{
w+=txd(A)*x;
}
if (twist==MR_SEXTIC_M)
{
w+=tx(B);
}
if (twist==MR_SEXTIC_D)
{
w+=txd(B);
}
if (twist==MR_QUADRATIC)
{
w+=tx(tx(A))*x+tx(tx(tx(B)));
}
}
else
{
w+=A*x+B;
}
return w;
}
BOOL is_on_curve(const ZZn4& x)
{
ZZn4 w;
w=rhs(x);
if (qr(w)) return TRUE;
return FALSE;
}
BOOL qr(const ZZn4& x)
{
ZZn2 a,s;
if (x.iszero()) return TRUE;
x.get(a,s);
if (s.iszero()) return TRUE;
s*=s;
a*=a; a-=txx(s);
if (!qr(a)) return FALSE;
return TRUE;
// s=sqrt(a);
// if (qr((x.a+s)/2) || qr((x.a-s)/2)) return TRUE;
// return FALSE;
}
ZZn4 sqrt(const ZZn4& x)
{
// sqrt(a+xb) = sqrt((a+sqrt(a*a-n*b*b))/2)+x.b/(2*sqrt((a+sqrt(a*a-n*b*b))/2))
// sqrt(a) = x.sqrt(a/n)
// where x*x=n
ZZn4 w;
ZZn2 a,s,t;
if (x.iszero()) return w;
x.get(a,s);
if (s.iszero())
{
if (qr(a))
{
s=sqrt(a);
w.set(s,0);
}
else
{
s=sqrt(txd(a));
w.set(0,s);
}
return w;
}
s*=s;
a*=a; a-=txx(s);
s=sqrt(a);
if (s.iszero()) return w;
x.get(t);
if (qr((ZZn2)((t+s)/2)))
{
a=sqrt((t+s)/2);
}
else
{
a=sqrt((t-s)/2);
if (a.iszero()) return w;
}
x.geth(t);
w.set(a,t/(2*a));
return w;
}
ZZn4 conj(const ZZn4& x)
{
ZZn4 u;
zzn4_conj((zzn4 *)&x.fn,(zzn4 *)&u.fn);
return u;
}
// For use by ZZn8
ZZn4 tx(const ZZn4& x)
{
ZZn4 w=x;
zzn4_tx(&w.fn);
return w;
}
ZZn4 txd(const ZZn4& x)
{
ZZn2 u,v;
x.get(u,v);
u=txd(u);
ZZn4 w(v,u);
return w;
}
// ZZn4 powering of unitary elements
ZZn4 powu(const ZZn4& x,const Big& k)
{
int i,j,nb,n,nbw,nzs;
ZZn4 u,u2,t[11];
Big k3;
if (k==0) return (ZZn4)one();
u=x;
if (k==1) return u;
//
// Prepare table for windowing
//
k3=3*k;
u2=(u*u);
t[0]=u;
for (i=1;i<=10;i++)
t[i]=u2*t[i-1];
nb=bits(k3);
for (i=nb-2;i>=1;)
{
n=naf_window(k,k3,i,&nbw,&nzs,11);
for (j=0;j<nbw;j++) u*=u;
if (n>0) u*=t[n/2];
if (n<0) u*=conj(t[(-n)/2]);
i-=nbw;
if (nzs)
{
for (j=0;j<nzs;j++) u*=u;
i-=nzs;
}
}
return u;
}
// regular ZZn4 powering - but see powl function in zzn2.h
ZZn4 pow(const ZZn4& x,const Big& k)
{
int i,j,nb,n,nbw,nzs;
ZZn4 u,u2,t[16];
if (k==0) return (ZZn4)1;
u=x;
if (k==1) return u;
//
// Prepare table for windowing
//
u2=(u*u);
t[0]=u;
for (i=1;i<16;i++)
t[i]=u2*t[i-1];
// Left to right method - with windows
nb=bits(k);
if (nb>1) for (i=nb-2;i>=0;)
{
n=window(k,i,&nbw,&nzs,5);
for (j=0;j<nbw;j++) u*=u;
if (n>0) u*=t[n/2];
i-=nbw;
if (nzs)
{
for (j=0;j<nzs;j++) u*=u;
i-=nzs;
}
}
return u;
}
// standard MIRACL multi-exponentiation
ZZn4 pow(int n,const ZZn4* x,const Big* b)
{
int k,j,i,m,nb,ea;
ZZn4 *G;
ZZn4 r;
m=1<<n;
G=new ZZn4[m];
for (i=0,k=1;i<n;i++)
{
for (j=0; j < (1<<i) ;j++)
{
if (j==0) G[k]=x[i];
else G[k]=G[j]*x[i];
k++;
}
}
nb=0;
for (j=0;j<n;j++)
if ((k=bits(b[j]))>nb) nb=k;
r=1;
for (i=nb-1;i>=0;i--)
{
ea=0;
k=1;
for (j=0;j<n;j++)
{
if (bit(b[j],i)) ea+=k;
k<<=1;
}
r*=r;
if (ea!=0) r*=G[ea];
}
delete [] G;
return r;
}
ZZn4 powl(const ZZn4& x,const Big& k)
{
ZZn4 w8,w9,two,y;
int i,nb;
two=(ZZn)2;
y=two*x;
if (k==0) return (ZZn4)two;
if (k==1) return y;
w8=two;
w9=y;
nb=bits(k);
for (i=nb-1;i>=0;i--)
{
if (bit(k,i))
{
w8*=w9; w8-=y; w9*=w9; w9-=two;
}
else
{
w9*=w8; w9-=y; w8*=w8; w8-=two;
}
}
return (w8/2);
}
ZZn2 real(const ZZn4 &x)
{
ZZn2 r;
x.get(r);
return r;
}
ZZn2 imaginary(const ZZn4 &x)
{
ZZn2 i;
x.geth(i);
return i;
}
#ifndef MR_NO_STANDARD_IO
ostream& operator<<(ostream& s,const ZZn4& xx)
{
ZZn4 b=xx;
ZZn2 x,y;
b.get(x,y);
s << "[" << x << "," << y << "]";
return s;
}
#endif