Cryptography and Number Theory

Transcription

1 Chapter 2 Cryptography and Number Theory 2.1 Cryptography and Modular Arithmetic Introduction to Cryptography For thousands of years people have searched for ways to send messages in secret. For example, in ancient times a king desired to send a secret message to his general in battle. The king took a servant, shaved his head, and wrote the message on his head. He then waited for the servant s hair to grow back and then sent the servant to the general. The general then shaved the servant s head and read the message. If the enemy had captured the servant, they presumably would not have known to shave his head, and the message would have been safe. Cryptography is the study of methods to send and receive secret messages; it is also concerned with methods used by the adversary to decode messages. In general, we have a sender who is trying to send a message to a receiver. There is also an adversary, who wants to steal the message. We are successful if the sender is able to communicate a message to the receiver without the adversary learning what that message was. In traditional cryptography, the sender and receiver agree in advance on a secret code, and then send messages using that code. For example, one of the oldest types of code is known as a Caesar cipher. In this code, the letters of the alphabet are shifted by some fixed amount. Typically, we call the original message the plaintext and the encoded text the ciphertext. Then an example of a Caesar cipher would be the following code plaintext A B C D E F G H I J K L M N O P Q R S T U V W X Y Z ciphertext E F G H I J K L M N O P Q R S T U V W X Y Z A B C D Thus if we wanted to send the plaintext message we would send the ciphertext ONE IF BY LAND AND TWO IF BY SEA SRI MJ FC PERH ERH XAS MJ FC WIE 25

2 26 CHAPTER 2. CRYPTOGRAPHY AND NUMBER THEORY Note that it is easy for the sender to encode the message, and for the receiver to decode the message. The disadvantage of this scheme is that it is also easy for the adversary to just try the 26 different possible Caesar ciphers and decode the message (only one will decode into plain English, with very high probability). Of course, there is no reason to use such a simple code; we can use any arbitrary permutation of the alphabet as the ciphertext, e.g. plaintext A B C D E F G H I J K L M N O P Q R S T U V W X Y Z ciphertext H D I E T J K L M X N Y O P F Q R U V W G Z A S B C If we encode a short message with a code like this, it would be hard for the adversary to decode it. However, with a message of any reasonable length (greater than about 50 letters), an adversary with a knowledge of the statistics of the English language can easily crack the code. There is no reason that we have to use simple mappings of letters to letters. For example, our coding algorithm can be to take three consecutive letters, reverse their order, interpret each as a base 26 integer (with A=0;B=1, etc.) multiply that number by 37, add 95 and then convert that number to base 8. Continue with each three letters, appending as we go along, using 8 s and 9 s to separate the numbers. When we are done, we reverse the number, and replace each digit 5 by two 5 s. Here is an example of this method: plaintext: ONEIFBYLANDTWOIFBYSEA block and reverse: ENO BFI ALY TDN IOW YBF AES base 26 integer: * base 8: appended : reverse, 5rep : Now, we can verify that it is possible for the receiver, who knows the code, to decode this message, and that the casual reader of the message will have no hope of decoding it. So it seems that with a complicated enough code, we can have secure cryptography. Unfortunately, there are at least two flaws with this method. The first is that if the adversary learns, somehow, what the code is, then she can easily decode it. Second, if this coding scheme is repeated often enough, and if the adversary has enough time, money and computing power, this code could be broken. In the field of cryptography, there are definitely players who have all these resources (such as a government, or a large corporation). The infamous German Enigma is an example of a much more complicated coding scheme, yet it was broken and this helped the Allies win World War II. (The reader might be interested in looking up more details on this.) In general, any scheme that uses a codebook, a secretly agreed upon (possibly complicated) code, suffers from these drawbacks. These codes are also called one-time pads, as the communicating parties agree on using the same pad or page in a codebook. A public-key cryptosystem overcomes the problems mentioned above. In a public key cryptosystem, the sender and receiver (often called Alice and Bob respectively) don t have to agree in advance on a secret code. In fact, they each publish part of their code in a public directory. Further, the adversary can intercept the message, and can have the public directory, and these will not be able to help him decode the message.

3 2.1. CRYPTOGRAPHY AND MODULAR ARITHMETIC 27 More precisely, Alice and Bob will each have two keys, a public key and a secret key. W e will denote Alice s public and secret keys as P A and S A and Bob s as P B and S B. They each keep their secret keys to themselves, but can publish their public keys and make them available to anyone, including the adversary. For convenience, we will think of these keys as functions; in particular we require that they each be one-to-one functions from the set D of possible messages onto itself. The public and secret keys are chosen to be inverses of each other, i.e for any M D we have that M = S A (P A (M)) = P A (S A (M)). (2.1) We also assume that, for Alice, S A and P A are easily computable. However, it is essential that for everyone except Alice, S A is hard to compute, even if you know P A. At first glance, this may seem to be an impossible task, Alice creates a function P A, that is public and easy to compute for everyone, yet this function has an inverse, S A, that is hard to compute for everyone except Alice. It is not at all clear how to design such a function. The first such cryptosystem is the now-famous RSA cryptosystem, widely used in many contexts. To understand how such a cryptosystem is possible will rely on some knowledge of number theory and on computational complexity. We will develop the necessary number theory in the next few sections. Before doing so, let us just assume that we have such a function and see how we can make use of it. Suppose Alice wants to send Bob a message M, she takes the following two steps: Alice obtains Bob s public key P B. Alice applies Bob s public key to M to create ciphertext C = P B (M). Alice then sends C to Bob. Bob can decode the message by using his secret key to compute S B (C) which is identical to S B (P B (M)), which by (2.1) is identical to M the original message. The beauty of the scheme is that even if the adversary has C and knows P B, she cannot decode the message without S B.ButS B is a secret that only Bob has. And even though the adversary knows that S B is the inverse of P B, the nature of P B is such that the adversary cannot easily compute its inverse Congruence modulo m The RSA encryption scheme is built upon the idea of congruence mod n. The symbolism m mod n means the remainder we get when we divide m by n. A bit more precisely, for integers m and n, m mod n is the smallest nonnegative integer r such that for some integer q. 1 m = nq + r (2.2) 1 In an unfortunate historical evolution of terminology, the fact that for every nonnegative integer m and positive integer n, there exist unique nonnegative integers q and r such that m = nq + r and r<nis called Euclid s algorithm. In modern language we would call this Euclid s Theorem instead. While it seems obvious that there is such a smallest nonnegative integer r and that there is exactly one such pair q, r with r<n, a technically complete study would derive these facts from the basic axioms of number theory, just as obvious facts of geometry are derived form the basic axioms of geometry. The reasons why mathematicians take the time to derive such obvious facts from basic axioms is so that everyone can understand exactly what we are assuming as the foundations of our subject; as the rules of the game in effect.

4 28 CHAPTER 2. CRYPTOGRAPHY AND NUMBER THEORY Compute 10 mod 7, -10 mod 7, 10 mod 7. The last one has two possible interpretations. We are looking for the one that is an integer Compute 21 mod 9, 38 mod 9, mod 9, (21 mod 9) (38 mod 9), mod 9, (21 mod 9) + (38 mod 9), 21 mod mod Show that 16 mod 9 = 34 mod 9. Is the same true of the pairs 16+1 and 34+1? 16+2 and 34+2? 16+7 and 34+7? What about 16 + k and 34 + k for an arbitrary integer k? In Exercise 2.1-1, 10 mod 7 is 3, while 10 mod 7 is 4. Note that 3 mod 7 is 4 also. Also, mod 7 = 0, suggesting that -10 is essentially the same as -3 when we are considering integers mod 7. We will be able to make this idea more precise in a few paragraphs. What do we mean by 1 10 mod 7? In algebra, 1 10 means the number that we multiply by 10 in order to get one. If we try multiplying ten by the numbers 1,2,3,4,5,6,etc., and taking the result mod n, weseethat5 10 = 50 and 50 mod 7 = 1, so that so it is natural to say that 1 10 mod 7 = 5. In Exercise 2.1-2, the point to notice is that and 5 10 mod 7 = 1, mod 9 = (21 mod 9)(38 mod 9) mod 9 = (21 mod 9) + (38 mod 9). These equations are very suggestive, though the general equations that they first suggest aren t true! Some closely related equations are true; the idea of congruence modulo n that we explore below allows us to discover these equations. Closely related to the operation m mod n is the relationship of congruence modulo n. Wesay that p and q are congruent modulo n if p mod n = q mod n. Another way to say this is to say that p and q are congruent mod n if p q is an integer multiple of n. The usual way of writing the statement that p and q are congruent mod n is p q (mod n). The congruence class of m mod n, denoted by m when the n is clear from context, is the set of all numbers congruent to m mod n. Thus every number is in some congruence class mod n. Two different congruence classes mod n cannot have any numbers in common, because if k were both i and j, then i mod n = k mod n = j mod n, so that i and j are congruent. However i and j are arbitrary members of their classes, so what we can conclude is that every member of i is congruent to every member of j. Thus the classes i and j have exactly the same members and are thus the same set. Therefore different congruence classes must be disjoint. This proves our next theorem: Theorem Congruence mod n is an equivalence relation.

5 2.1. CRYPTOGRAPHY AND MODULAR ARITHMETIC 29 Proof: Given above. Exercise is really a statement about congruence modulo n. All the pairs given in Exercise have the same mods, and the last pair corresponds to the fact that if i j (mod n), then for any integer k, i + k j + k (mod n). While we can think about (and define rigorously) modular arithmetic via congruences and congruence classes, a slightly different approach will turn out to be more closely related to what we do computationally. We will pull the smallest positive number out of each congruence class, and use that instead of the class itself. Thus, if we are doing arithmetic mod n, we will restrict ourselves to using the numbers 0, 1,...,n 1. We will redefine addition to be addition modulo n, and multiplication to be multiplication modulo n. We will use the notation Z n to represent the integers 0, 1,...,n 1 together with a redefinition of addition, which we denote by + n,and a redefinition of multiplication, which we denote n. The redefinitions are: i + n j = (i + j) modn (2.3) i n j = (i j) modn (2.4) We must now verify that all the usual rules of arithmetic normally apply to addition and multiplication still apply with + n and n. In particular, we wish to verify the commutative, associative and distributive laws. The intuition behind why these laws will hold is that as we are doing addition and multiplication on integers, if the final result is going to be expressed mod n, then at any point in the computation, we can take any partial sum or product mod n. Thiscan be formalized in the following lemma. Lemma (i + j) modn =(i mod n + j mod n) modn. (2.5) Proof: Using (2.2), we can write i as k 1 n + r 1,wherer 1 <n,andcanwritej as k 2 n + r 2, where r 2 <n. Then we can rewrite the left side of (2.5) as (i + j) modn = (k 1 n + r 1 + k 2 n + r 2 )modn = ((k 1 + k 2 )n +(r 1 + r 2 )) mod n = (r 1 + r 2 )modn, where the last line follows because any integral multiple of n is 0 mod n. Wecanrewritethe right side of (2.5) as (i mod n + j mod n) modn = ((k 1 n + r 1 )modn +(k 2 n + r 2 )modn) modn = (r 1 mod n + r 2 mod n) modn = (r 1 + r 2 )modn, where the last line follows because r 1 <nand r 2 <n. Similarly, we can show that (i + j) modn =(i + j mod n) modn =(i mod n + j) modn. (2.6) We now turn to verifying the various laws of addition and multiplication. First, the fact that + n is commutative follows immediately from the definition. We will now prove that it is also associative.

6 30 CHAPTER 2. CRYPTOGRAPHY AND NUMBER THEORY Lemma For any x, y, z Z n, (x + n y)+ n z = x + n (y + n z) Proof: following We will repeatedly use the definition of + n and equations 2.5 and 2.6 to get the (x + n y)+ n z = ((x + n y)+z) modn = ((x + n y)+z mod n) modn = ((x + y) modn + z mod n) modn = ((x mod n + y mod n) modn + z mod n) modn = (x mod n + y mod n + z mod n) modn = (x mod n +(y mod n + z mod n) modn) modn = (x mod n +(y + z) modn) modn = (x +(y + z) modn) modn = (x + y + n z)modn = x + n (y + n z). We can also show that a version of Equation 2.5 and Equation 2.6 holds with addition replaced by multiplication, i.e. that (i j) modn =(i mod n j mod n) modn =(i j mod n) modn =(i mod n j) modn. (2.7) The proof of this will be an exercise. Using this we can show that n is associative. (The proof of this is also an exercise.) Given these, we can show that the distributive law also holds, i.e. Lemma For any x, y, z Z n, x n (y + n z)=(x n y)+ n (x n z). Proof: Using (2.5), (2.6) and (2.7), x n (y + n z) = (x(y + n z)) mod n = (x((y + z) modn)) mod n = (x((y mod n + z mod n) modn)) mod n = ((x mod n)((y mod n + z mod n) modn)) mod n = ((x mod n)(y mod n + z mod n)) mod n = ((x mod n)(y mod n)+(x mod n)(z mod n)) mod n = (((x mod n)(y mod n)) mod n +((x mod n)(z mod n)) mod n) modn = ((xy) modn +(xz) modn) modn = ((x n y)+(x n z)) mod n = (x n y)+ n (x n z))

7 2.1. CRYPTOGRAPHY AND MODULAR ARITHMETIC 31 We conclude this section by observing that repeated applications of Lemma are useful when computing sums in which the numbers are large. For example, suppose you had m integers x 1,...,x m and you wanted to compute m j=1 x i mod m. One natural way to do so would be to compute the sum, and take the result modulo m. However, it is possible that, on the computer that you are using, even though m j=1 x i mod m is a number that can be stored in an integer, and each x i can be stored in an integer, m j=1 x i mightbetoolargetobestoredinaninteger. (Recall that integers are typically stored as 4 or 8 bytes, and thus have a maximum value of roughly or ) Another way to say this is that if we are computing a result mod n, we should do all our calculations in Z n using + n and n. Cryptography Revisited One natural way to use addition of a number m mod n in encryption is to first convert the message to a sequence of digits say concatenating all the ASCII codes for all the symbols in the message and then simply add m to the message mod n. Ifn happens to be larger than the message in numerical value, then it is simple for someone who knows m to decode the encrypted message. However an adversary who sees the encrypted message has no special knowledge and so unless m was ill chosen (for example m = 1 would be a silly choice) the adversary who knows what system you are using, even including the value of n, but does not know m, is essentially reduced to trying all possible m values. This is an example of a one time pad, discussed in Section 2.1.1, because if you use m only once, there is no way for the adversary to collect any data that will aid in guessing m. Thus, if only you and your intended recipient know m, this kind of encryption is quite secure: guessing m is just as hard as guessing the message. It is possible that once n has been chosen, you will find you have a message which translates to a larger number than n. Normally you would then break the message into segments, each with no more digits than n, and send the segments individually. So long as you were not sending a large number of segments, it would still be quite difficult for your adversary to guess m by observing the encrypted information. In fact, it would not be unusual to use the same value of m for a day, say, (or for one session on the internet) and then to change it. However, if you used the value often enough, the adversary could collect enough information to be able to break the code Multiplication Mod n Show that (i j) modn =(i mod n)(j mod n) modn Would it make sense to encode a message by multiplying it by a key mod n? When we encoded a message by adding m mod n, we could decode the message simply by subtracting m mod n. By analogy, if we encode by multiplying by m mod n, we would expect to decode by dividing by m mod n. However division mod n can be problematic for some values of n. Let us do a trivial example. Suppose your value of n was 12 and the value of m was 4. You send the message 3 as (mod 12). Thus you send the encoded message 0. Now your partner sees 0, and says the message might have been 0; after all, (mod 12). On the

8 32 CHAPTER 2. CRYPTOGRAPHY AND NUMBER THEORY other hand, (mod 12), (mod 12), and (mod 12) as well. Thus your partner has four different choices for the original message, which is almost as bad as having to guess the original message itself! You might think that 0 would present special problems for division, so let s look at another example. Suppose that m =3andn = 12. Now we encode the message 6 by computing (mod 12). Simple calculation show that (mod 12), (mod 12), and (mod 12). Thus in this case, there are three possible ways to decode the message. Let s look at one more example that will provide some hope. Let m =5andn = 12, and let s encode the message 7 as (mod 12). This time, simple calculation shows that 7 is the unique solution (mod 12) to the equation x 5 11 (mod 12). Thus in this case we can correctly decode the message. As we shall see in the next section, the kind of problem we had happens only when m is a factor of n. Thus all our receiver needs to know, in our cases, is how to divide by m mod n, and she can decrypt our message. If you don t now know how to divide by m, then you can begin to understand the idea of public key cryptography. The message is there for anyone who knows how to divide by m to find, but if nobody but our receiver can divide by m, we can tell everyone what m is and our messages will still be secret. As we shall soon see, dividing by m is not particularly difficult, so a better trick is needed for public key cryptography to work. Exercises E2.1-1 State and prove the commutative and associative law for n. E2.1-2 It is obvious how to solve any equation of the form x + n a = b in Z n, and that the result will be a unique value of x. On the other hand we saw that 0, 3, 6, and 9 are all solutions to the equation 4 n x =0 in Z 12. Are there any equations of the form a n x = b that don t have any solutions at all in Z 12? If so, give one, and find out whether there are any numbers a such that each equation of the form a n x = b does have a solution. On the other hand if each equation of the form a n x = b has a solution in Z 12, give a proof of this. (Note that having a solution is different from having a unique solution.) E2.1-3 Does every equation of the form a n x = b have a solution in Z 5?inZ 7?inZ 9?in Z 11? E2.1-4 Recall that if a prime number divides a product of two integers, then it divides one of the factors. a) Use this to show that as b runs though the integers from 0 to p 1, with p prime, the products a n b are all different (for each fixed choice of a between 1 and p 1). b) Explain why every integer greater than 0 has a unique multiplicative inverse mod p, ifp is prime.

9 2.1. CRYPTOGRAPHY AND MODULAR ARITHMETIC 33 E2.1-5 Modular arithmetic is used in generating psuedo-random numbers. One basic algorithm (still used on many machines!) is by linear congruential random number generation. The following piece of code generates a sequence of numbers that may appear random to an unaware user. seed := x := 0 /* set seed */ repeat x := (ax + b) modn print x until x = seed Execute the loop for a =3,b=7,n = 11. E2.1-6 Write down the 7 multiplication table for Z 7.