Skip to content

Hyperparameter Optimization of Tree Parity Machines to Minimize the Effectiveness of Unconventional Attacks on Neural Cryptography​.

Notifications You must be signed in to change notification settings

MLTransfer/MLEncrypt-Research

Repository files navigation

MLEncrypt-Research

Usage

  • single: runs a single instance of neural key exchange
  • multiple: runs many instances of neural key exchange, all with the same configuration (this can be useful for benchmarking)
  • hparams: runs hyperparameter optimization (different instances of neural key exchange will likely have different configurations)

Note: It may appear that enabling JIT compilation results in a slowdown. If you don't want to run with JIT, then don't pass the XLA flags to the run command.

Note: Our script features very verbose logging to TensorBoard. Enable this by passing -tb to the run command.

Note: You can choose to not calculate the synchronization score for each iteration. Enable this by passing -b to the run command.

CPU

TF_XLA_FLAGS="--tf_xla_auto_jit=2 --tf_xla_cpu_global_jit" poetry run mlencrypt-research single

GPU

TF_XLA_FLAGS=--tf_xla_auto_jit=2 poetry run mlencrypt-research single

Acknowledgements

About

Hyperparameter Optimization of Tree Parity Machines to Minimize the Effectiveness of Unconventional Attacks on Neural Cryptography​.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages