Exponentiation ciphers are due to Pohlig and Hellman (1978). They are less vulnerable to frequency analysis than block ciphers. Here's the procedure.
1. Let p be a prime number, and let e be the exponent. They should satisfy .
2. Encode the letters of the alphabet as
3. Group the letters in the message in blocks of m letters, where m is chosen so that
For example, suppose . Then you should use blocks of letters, because . And if , you should use blocks of letters, because . This stipulation merely ensures that the blocks are unique mod p.
4. Encode a block A using
The ciphertext C is an integer satisfying and this integer is the ciphertext: You don't convert it to letters.
Example. Let , so . Take an exponent relatively prime to 2620 --- for instance, .
I use blocks of letters, because
Take the plaintext and convert it to numbers:
Now encode the message:
The ciphertext is
How should you do these computations? The best way to do the computations is to use software which can do large-integer arithmetic. Most calculators can only accomodate 10--20 digit integers. If you try to compute on a calculator, you'll find that it's around . Because these computations require modular arithmetic, you can't use floating point --- you are losing significant digits.
So how do you do something like if all you have is a calculator? First, rewrite it:
Now I'll compute and reduce it mod 2621:
(I got the last result by finding . Subtract the integer part (2223) times 2621 from 5827396: .)
Therefore,
It should be clear how to proceed. Use the rules for exponents to reduce the product a little bit at a time, so that the intermediate results don't overflow your calculator.
Obviously, it is easier to use a computer!
To decode a message that has been encoded using an exponentation cipher, find d such that
This is possible (using the Euclidean algorithm), since by assumption. Equivalently, for some k. Now suppose . Then
Note that A is less than (m 25's) because A came from a block of m letters. Since , it follows that , and little Fermat applies. Thus, .
In other words, raising C to the d-th power recovers the plaintext from the ciphertext.
Example. Take and . ; apply the Extended Euclidean algorithm:
Hence, .
So to decode , raise it to the 1191-th power:
, which is the plaintext for this block.
In a public-key cryptosystem, there are separate keys for encoding and decoding messages. One key is public, so that anyone can send a message to me. But I'm the only one who knows the private key, so I'm the only one who can read my messages. Moreover, I can use my private key to send messages, which can be decoded using the public key. Since I'm the only one who could have encoded such a message, people know the message must have come from me --- a digital signature.
I'll discuss the RSA public-key cryptosystem, which is due to Rivest, Shamir, and Adleman (1978). You'll see that it's essentially a modified exponentiation cipher.
1. Let p and q be large prime numbers. For practical applications, you'll need primes which are around 100 digits long. Let . (n is called the key.)
2. Find an exponent e such that , and such that .
If n were prime, would be , and I'd have the setup for an exponentiation cipher. The condition guarantees that you can't recover the plaintext A by taking e-th roots. For if A is any block besides or , the result is when it's raised to the e-th power, so it changes when it's reduced mod n.
3. Encode the letters of the alphabet as
4. Group the letters in the message in blocks of m letters, where m is chosen so that
5. Encode a block A using
Example. Let . Then
I can choose e to be any number relatively prime to 2520, and such that . I'll take .
Since , I use blocks of two letters.
Take the plaintext and convert it to numbers:
Now encode the message:
The ciphertext is
When this system is used, e and n are made public so people can encipher messages. The security of this method depends on the difficulty of finding , since (as I'll show below) this is what you need to decode a message.
On the one hand, if you know p and q, then
Since p and q are known, so is .
On the other hand, suppose you know , you don't know p and q, but you do know that n is a product of two primes p and q. Then
Therefore,
Moreover,
The last two equations show that if you know (and n), then you can find , and from that you can find . But
Thus, you know p and q.
To summarize, knowing is equivalent to knowing p and q.
If p and q are 100-digit primes, then is around 200 digits. With present technology, it's hard to factor an arbitrary 200-digit number. It follows that finding --- and hence, breaking the code --- is difficult at the moment, which means the system is fairly secure.
Of course, no cipher is immune to human carelessness! If you let someone discover your key, the cipher is worthless.
Now here's how knowing allows you to decode a message. The idea is similar to that used in the exponentiation cipher.
Find d such that
This is possible (using the Euclidean algorithm), since by assumption. Equivalently, for some k. Now suppose . Then
is a consequence of Euler's theorem, and will be true provided . Now , so it's possible for this to fail if the plaintext A has either p or q as a prime factor. However, if p and q are each around 100 digits long, the probability that this will happen is around --- so it's nothing to worry about.
Just as in the exponentiation cipher, raising C to the d-th power recovers the plaintext from the ciphertext.
Example. Take and . I'll show that 2114 is not an enciphered message by decoding it. Recall that . Apply the Extended Euclidean algorithm:
Then
However, 80 can't be a block in a message, because it's greater than 25. Therefore, 2114 is not a ciphertext for this key.
Copyright 2007 by Bruce Ikenaga