Cryptanalysis of Keystream Generators using Machine Learning Techniques

Requirements

• python 3.6 (the version has to be exactly 3.6.*, since tensorflow doesn't support 3.7 yet)
• virtualenv (very useful but not necessary)

Installation

$ git clone https://github.com/luukbox/thesis-implementation.git
$ cd thesis-implementation
# if you want to work with a virtual environment run the next 2 commands
# otherwise skip to pip install
$ virtualenv -p /path/to/the/python/bin venv
$ source venv/bin/activate # activate the virtual environment
$ python --version # should be 3.6.* at this point !!!!
$ pip install -r requirements.txt # install the dependencies

Generate the binary sequences

1. uncomment generate_r_sequence() in generate_source_files.py
2. run python generate_source_files.py in the console
3. revert step 1.
4. uncomment generate_pr_sequence(num_bits=960000, poly=[11, 9]) and change the poly if you like to
5. step 2.

Test the binary sequence

1. paste the path into test_sequence.py
2. run test_sequence.py . The sequence is tested against an implementation of the NIST SP800-22 Rev 1a PRNG test suite

Merge binary sequences

Random.org's quota is 1M bits per day. To create larger datasets, we need to merge the random sequences.

$ python merge_binary_sequences.py path_to_seq1 path_to_seq2 output_name

The merged sequence will be saved into the binary_sequences folder and postfixed with the amount of bits it contains.

Generate a dataset

1. paste the paths of the binary sequences into generate_dset.py (random_raw_path="", pseudorandom_raw_path=""), change the NAME and other parameters e.g. the input_size
2. run python generate_dset.py in the console

Train a model

1. paste the dataset_path into train.py
2. run python train.py in the console