RC5

The key expansion algorithm is illustrated below, first in pseudocode, then example C code copied directly from the reference paper's appendix.

Following the naming scheme of the paper, the following variable names are used:

• w - The length of a word in bits, typically 16, 32 or 64. Encryption is done in 2-word blocks.
• u = w/8 - The length of a word in bytes.
• b - The length of the key in bytes.
• K[] - The key, considered as an array of bytes (using 0-based indexing).
• c - The length of the key in words (or 1, if b = 0).
• L[] - A temporary working array used during key scheduling. initialized to the key in words.
• r - The number of rounds to use when encrypting data.
• t = 2(r+1) - the number of round subkeys required.
• S[] - The round subkey words.
• P_w - The first magic constant, defined as ${\displaystyle Odd((e-2)*2^{w})}$, where Odd is the nearest odd integer to the given input, e is the base of the natural logarithm, and w is defined above. For common values of w, the associated values of P_w are given here in hexadecimal:
  • For w = 16: 0xB7E1
  • For w = 32: 0xB7E15163
  • For w = 64: 0xB7E151628AED2A6B
• Q_w - The second magic constant, defined as ${\displaystyle Odd((\phi -1)*2^{w})}$, where Odd is the nearest odd integer to the given input, where ${\displaystyle \phi }$ is the golden ratio, and w is defined above. For common values of w, the associated values of Q_w are given here in hexadecimal:
  • For w = 16: 0x9E37
  • For w = 32: 0x9E3779B9
  • For w = 64: 0x9E3779B97F4A7C15

# Break K into words
# u = w / 8
c = ceiling(max(b, 1) / u)
# L is initially a c-length list of 0-valued w-length words
for i = b-1 down to 0 do:
    L[i / u] = (L[i / u] <<< 8) + K[i]
     
# Initialize key-independent pseudorandom S array
# S is initially a t=2(r+1) length list of undefined w-length words
S[0] = P_w
for i = 1 to t-1 do:
    S[i] = S[i - 1] + Q_w
    
# The main key scheduling loop
i = j = 0
A = B = 0
do 3 * max(t, c) times:
    A = S[i] = (S[i] + A + B) <<< 3
    B = L[j] = (L[j] + A + B) <<< (A + B)
    i = (i + 1) % t
    j = (j + 1) % c

# return S

The example source code is provided from the appendix of Rivest's paper on RC5. The implementation is designed to work with w = 32, r = 12, and b = 16.

void RC5_SETUP(unsigned char *K)
{
   // w = 32, r = 12, b = 16
   // c = max(1, ceil(8 * b/w))
   // t = 2 * (r+1)
   WORD i, j, k, u = w/8, A, B, L[c];
   
   for (i = b-1, L[c-1] = 0; i != -1; i--)
      L[i/u] = (L[i/u] << 8) + K[i];
   
   for (S[0] = P, i = 1; i < t; i++)
      S[i] = S[i-1] + Q;
   
   for (A = B = i = j = k = 0; k < 3 * t; k++, i = (i+1) % t, j = (j+1) % c)
   {
      A = S[i] = ROTL(S[i] + (A + B), 3);
      B = L[j] = ROTL(L[j] + (A + B), (A + B));
   }
}

Encryption involved several rounds of a simple function. 12 or 20 rounds seem to be recommended, depending on security needs and time considerations. Beyond the variables used above, the following variables are used in this algorithm:

• A, B - The two words composing the block of plaintext to be encrypted.

A = A + S[0]
B = B + S[1]
for i = 1 to r do:
    A = ((A ^ B) <<< B) + S[2 * i]
    B = ((B ^ A) <<< A) + S[2 * i + 1]

# The ciphertext block consists of the two-word wide block composed of A and B, in that order.
return A, B

The example C code given by Rivest is this.

void RC5_ENCRYPT(WORD *pt, WORD *ct)
{
   WORD i, A = pt[0] + S[0], B = pt[1] + S[1];
   
   for (i = 1; i <= r; i++)
   {
      A = ROTL(A ^ B, B) + S[2*i];
      B = ROTL(B ^ A, A) + S[2*i + 1];
   }
   ct[0] = A; ct[1] = B;
}

Decryption is a fairly straightforward reversal of the encryption process. The below pseudocode shows the process.

for i = r down to 1 do:
    B = ((B - S[2 * i + 1]) >>> A) ^ A
    A = ((A - S[2 * i]) >>> B) ^ B
B = B - S[1]
A = A - S[0]

return A, B

The example C code given by Rivest is this.

void RC5_DECRYPT(WORD *ct, WORD *pt)
{
   WORD i, B=ct[1], A=ct[0];
   
   for (i = r; i > 0; i--)
   {
      B = ROTR(B - S[2*i + 1], A) ^ A;
      A = ROTR(A - S[2*i], B) ^ B;
   }
   
   pt[1] = B - S[1]; pt[0] = A - S[0];
}