Transcript Document
Algorithms Design and its
Applications
Algorithms with numbers
Back Ground
• Number theory was once viewed as a beautiful but
largely useless subject in pure mathematics.
• Today number- theoretic algorithms are used widely ,
due in part to the invention of cryptographic schemes
based on large prime numbers.
• The feasibility of these schemes rests on our ability to
find large primes easily, while their security rests on
our inability to factor the product of large primes .
Size of Cost
• In this lecture, a "large input" typically means an
input containing "large integers" rather than an input
containing "many integers".
• We measure the size of an input in terms of the
number of bits required to represent that input.
• An algorithm with integer inputs a1, a2, ..., ak is a
polynomial - time algorithm if it runs in time
polynomial in lg a1, lg a2, ..., lg ak, that is, polynomial
in the lengths of its binary- encoded inputs.
Basic arithmetic
Bases and logs
• How many digits are needed to represent the number
N>=0 in base b?
• How much does the size of a number change when
we change the base?
Addition
• The sum of any three single-digit numbers (with base
b>=2) is at most two digits long.
• Given two binary number x and y, how long does our
algorithm take to add them?
O(n), n is the number of bits of x and y.
Multiplication
• 13 * 11
Multiplication
• 乘法的时间复杂度是多少呢?
• 对于长度为n的乘数来说,将产生n个中间结果,
而对这些中间结果进行相加的次数是n-1次,从而
时间复杂度为O(n(n-1))=O(n2)
Multiplication
• another way to multiply
Multiplication
Multiplication
• Al Khwarizmi乘法算法的时间复杂度?
• 由于乘数每次都被取半,对于二进制来说,取半
意味着去掉最右边一位数,而乘数的长度为n,因
此该算法在递归n次后终结。每次递归需要进行一
次长度为n位的加法运算,其时间复杂度为O(n)。
因此Al Khwarizmi乘法算法的时间复杂度为O(n2)
。
Division
• To divide an integer x by another integer y≠0 means
to find a quotient q and a remainder r, where x = yq+r
and r <y.
Modular arithmetic
Modular Arithmetic Basic
Modular Arithmetic Basic
Modular Addition
• x+y mod N
• Since x and y are in the range 0 to N-1, their sum is
between 0 and 2(N-1). If the sum exceeds N-1,
subtract off N.
• The overall computation consists of an addition, and
possibly a subtraction, of numbers that never exceed
2N. the running time is O(n), where n = log N.
Modular multiplication
• xy mod N
• start with regular multiplication, then reduce the
answer modulo N. The product can be as large as (N1)2, at most 2n bits long since log(n-1)2 = 2log(N1)≤2n.
• The running time is O(n2).
Modular exponentiation
• 在密码学中,常需计算xy mod N. 这个的x,y和N均
为几百位长的整数。如何快速计算?
• 直接算xy,运算结果很大!即便x和y只有20位长
, xy也要大概1千万位长。
• 为保证中间运算结果不要太大,每步运算都模N.
Modular exponentiation
• Simple idea: repeatedly multiplying by x modulo N.
problem: if y is 500 bits long, we need to perform
y -1 ≈ 2500 multiplications!
Modular exponentiation
• better idea: starting with x and squaring repeatedly
modulo N, we get
we need to perform log y multiplications, ach takes
O(log2N) to compute.
• To determine xy mod N, we simply multiply together
and appropriate subset of these powers, those
corresponding to 1’s in the binary representation of y.
• A polynomial-time algorithm is within reach!
Modular exponentiation
sicily 1294. 高级机密
• 信息加密。
• 目前比较流行的编码规则称为RSA,是由美国麻
省理工学院的三位教授发明的。这种编码规则是
基于一种求密取模算法的:对于给出的三个正整
数a,b,c,计算a的b次方除以c的余数。
• 题目要求:计算 ab mod c
sicily 1294. 高级机密
问题分析
• 不好的算法:
– 先求出a的b次方,再模c。但题目给出的a,b,c的
范围比较大,要算出ab要用到高精度乘法,然
后模c还要用到高精度除法;
• 较好的算法:
– 利用同余的性质,
xy mod c = x * (y mod c) mod c
sicily 1294. 高级机密
代码
d = 1;
for (i = 1; i <= b; ++i) {
d = d * a % c;
}
cout << d;
Euclid’s algorithm for
greatest common divisor
• Euclid’s rule
If x and y are positive integer with x≥y, then gcd(x,y)=
gcd(x mod y, y)
• Proof.
因为 gcd(x,y)能整除x和y,因此整除x-y,即是x-y
的因子,因此gcd(x,y)≤ gcd(x-y, y).
而反过来推,同理可得gcd(x-y, y) ≤ gcd(x,y)。
故gcd(x,y)=gcd(x-y, y)。由此显然可得结论。
Euclid’s algorithm for
greatest common divisor
Euclid’s algorithm for
greatest common divisor
• This means that after any two consecutive round,
both a and b, are at the very least halved in value –
the length of each decreases by at least one bit. If they
are initially n-bits integers, then the base case will be
reached within 2n recursive calls. And since each call
involves a quadratic-time division, the total time is
O(n3)
An extension of Euclid algorithm
只要找到两个整数x和y,使得ax+by=d,且d是a和b
的因子,则d就是a和b的最大公因子;如果d是a和b
的最大公因子,则d一定可以表示为ax+by形式。
只要对欧几里得算法稍加扩展,即可找到所需的系
数x和y。
An extension of Euclid algorithm
An extension of Euclid algorithm
• Lemma
For any positive integers a and b, the extended Euclid
algorithm returns integers x, y, and d such that gcd(a,b) = d =
ax+by
• Proof.
对b做归纳假设。当b=0, 验证可知算法正确。
算法调用gcd(b,a mod b)来计算gcd(a,b)。
由于a mod b < b, 由归纳假设知返回结果是正确的.
Modular division
减法求最大公约数
于大整数而言,取模运算(其中用到除法)是非常昂贵的
开销,将成为整个算法的瓶颈。有没有办法能够不用取模
运算呢?
如果一个数能够同时整除x和y,则必能同时整除x-y和y;
而能够同时整x-y和y的数也必能同时整除x和y,即x和y的
公约数与x-y和y的公约数是相同的,其最大公约数也是相
同的,即f(x, y)= f(x-y, y),那么就可以不再需要进
行大整数的取模运算,而转换成简单得多的大整数的减法。
实例:f(42, 30)=f(30, 12)=f(12, 18)= f(18, 12)
= f(12, 6)= f(6, 6)= f(6, 0)= 6
不足之处。最大的瓶颈就是迭代的次数比之前的算法多了
不少,如果遇到(10 000 000 000 000, 1)
减法求最大公约数
代码
BigInt gcd(BigInt x, BigInt y)
{
if(x < y) return gcd(y, x);
if(y == 0) return x;
else return gcd(x - y, y);
}
求最大公约数算法三
算法一(欧几里得算法)的问题在于计算复杂的大整数除法
运算,而算法二虽然将大整数的除法运算转换成了减法运
算,降低了计算的复杂度,但它的问题在于减法的迭代次
数太多,如果遇到(10 000 000 000 000, 1)的情况就很
糟糕。
能否结合算法一和算法二从而使其成为一个最佳的算法呢?
求最大公约数算法三
记x和y的最大公约数为f(x, y)。
若x, y均为偶数,f(x, y)= 2 * f(x/2, y/2)= 2 * f
(x>>1, y>>1)
若x为偶数,y为奇数,f(x, y)= f(x/2, y)= f(x>>1,
y)
若x为奇数,y为偶数,f(x, y)= f(x, y/2)= f(x,
y>>1)
若x, y均为奇数,f(x, y)= f(x, x - y),
那么在f(x, y)= f(x, x - y)之后,(x - y)是一个偶
数,下一步一定会有除以2的操作。
最坏情况下的时间复杂度是O(log2(max(x, y))。
求最大公约数算法三
示例:
f(42, 30)= f(1010102, 111102)
= 2 * f(101012, 11112)
= 2 * f(11112, 1102)
= 2 * f(11112, 112)
= 2 * f(11002, 112)
= 2 * f(112, 112)
= 2 * f(02, 112)
= 2 * 112
=6
求最大公约数算法三
BigInt gcd(BigInt x, BigInt y)
{
if(x < y) return gcd(y, x);
if(y == 0) return x;
else {
if(IsEven(x)){
if(IsEven(y)) return (gcd(x >> 1, y >> 1) << 1);
else return gcd(x >> 1, y);
}
else {
if(IsEven(y)) return gcd(x, y >> 1);
else return gcd(y, x - y);
}
}
}
同余
• 同余
– 设m是正整数,a,b是整数,如果m|(a-b),则称a和b关于模m同余,
记作a≡b(mod m)或者说,如果a,b除以m的余数相等,则称a和b关
于模m同余
• 同余的性质
–
–
–
–
a≡a(mod m)
如果a≡b(mod m),则b≡a(mod m)
如果a≡b(mod m)且b≡c(mod m), a≡c(mod m)
如果a≡b(mod m)且c≡d(mod m),则a±c≡b± d(mod m),
ac≡bd(mod m)
同余
• 同余的性质(cont.)
–
–
–
–
–
如果a≡b(mod m),则an≡bn(mod m),n∈N
如果ac≡bc(mod m),则a≡b(mod (m/gcd(c,m))
如果a≡b(mod m)且d|m,则a≡b(mod d)
如果a≡b(mod m),则ad≡bd(mod m)
如果a≡b(mod mi),i=1,2,…,n,l=lcm(m1,m2,…,mn),则
a≡b(mod l)
– 如果p为素数,则ap ≡ a(mod p);如果gcd(a,p)=1,则ap-1 ≡
1(mod p)
Primality Testing
筛法求素数
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
筛法求素数
代码
const int MAX = 10000;
bool prime[MAX + 1];
void searchprime() {
memset(prime, true, sizeof(prime));
prime[1] = false;
for (int i = 2; i*i<= MAX; ++i)
{
if (prime[i]) {
int j = i * 2;
while (j <= MAX) {
prime[j] = false;
j += i;
}
}
}
}
代码(筛法求素数)
for (int i = 2; i <= (int) floor(sqrt(MAX)); ++i) {
if (prime[i]) {
int j = i * 2;
while (j <= MAX) {
prime[j] = false;
j += i;
}
}
}
}
Fermat’s little theorem
Fermat’s little theorem
• Proof.
Algorithm for testing primality
Algorithm for testing primality
Algorithm for testing primality
An algorithm for testing primality,
with low error probability
Carmichael numbers
• 561 = 3*11*17, not a prime.
• fool the Fermat test, because a560 ≡1 (mod 561) for all
values of a relatively prime to 561.
• Rabin and Miller algorithm.
Generating random primes
Generating random primes
• Q: what is the probability that the output of the
algorithm is really prime?
• A: suppose we perform the test with base a=2 for all
numbers N≤25∙109.
RSA
• RSA基本原理
• 选定一个数N,再选择一个N到N的双射函数f作为
加密密钥(公钥),该函数的逆函数作为解密密
钥(密钥)。f的选定必须使得其逆函数无法从f
推出。
RSA
RSA