About ChaC

We implemented a smaller version of ChaCha that uses 256 bits instead of 512 bits. We initially thought this would be more efficient because it's half the size, but we have since realized that it is equally as computationally expensive, as twice the number of blocks will need to be "shuffled".

Thus, we decided to approach the problem by reducing the average number of rounds used to generate a keystream.

Each block will generate 4 x 128-bit keystreams, instead of generating a new 128-bit key with each block. We hope that this change will significantly reduce the computation requirements of the algorithm while remaining relatively secure.

Our block is structured as follows: k is a 128-bit key, ct is a 32-bit counter, n is a 64-bit nonce, and C is a 32-bit constant, fixed to the value of 0x43686143.

---

k1	k2	k3	k4
ct	n1	n2	C

---

Here's a brief description of our new stream cipher.

We can generate a keystream using the following algorithm:

Let Round(a, n) represent the operation of running n rounds on the input a

Block 1 (256-bit matrix)

      k0 = Round(Block 1, 20)

      k1 = Round(k0, 2)

      k2 = Round(k1, 2)

      k3 = Round(k2, 2)

Block 2: (256-bit matrix)

      k4 = Round(Block 2, 20)

      k5 = Round(k4, 2)

      k6 = Round(k5, 2)

      k7 = Round(k6, 2)

      ... Then, we XOR the generated keystream with the input message to create the cypher text. The example above, consisting of two blocks, generates a 2 * 4 * 256 = 2048-bit keystream. A rough analysis shows that the above algorithm is at least 3 times as performant as ChaCha.

Usage:

Compile:

gcc chac.c -o chac

To change the key, change line 88.

Run the program:

./chac [-n nonce] input_file output_file

The program is symmetric. The encrypted file can be input to get the plaintext message.

NIST Testing

Create a keystream with 1-4M bits in binary.

wc -m sample.bin

1. Run the assessment with the desired chunk size. Depending on the test, chunk sizes can range from 256bits to 20,000. For ChaC we chose 20,000.

./assess <chunk size>

A chunk length of 20,000 is adequate with 200 chunks.

2. Enter 0 to select a file input.

3. Enter file name.

4. Do not run all tests. Enter 0.

5. Only run the following to prevent underflow on deep tests: Frequency, Block Frequency, Cumulative Sums, Runs, Longest Run, Approximate Entropy, Serial, Linear Complexity.

6. Enter the number of chunks.

7. Select binary as input

8. View test results in Experiments --> Algorithm Testing --> finalAnalysisReport.txt