New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Using external CMAES library #103

Open

Chrism7 wants to merge 14 commits into master from cma_hansen

Collaborator

Chrism7 commented Nov 23, 2017

We implemented a learner, experiment and console program concerning a reliability based CMAES attack on Arbiter PUFs. For optimality purposes we use an external library for applying a CMAES algorithm.
Up to now, only particular Arbiter PUFs can be learned.

nils-wisiol reviewed

View reviewed changes

pypuf/learner/evolution_strategies/reliability_based_cmaes.py Outdated

+                  @contextlib.contextmanager
+                  def avoid_printing():
+                      save_stdout = sys.stdout
+                      sys.stdout = open('trash', 'w')

Owner

nils-wisiol Dec 1, 2017

open('/dev/null', 'w')

nils-wisiol requested changes

View reviewed changes

Owner

nils-wisiol left a comment

A couple minor annotations in the code, but overall very nice!!

Also, what's up with test/test.py?

pypuf/experiments/experiment/reliability_based_cmaes.py Outdated

		@@ -0,0 +1,123 @@
		"""This module provides an experiment class which learns an instance
		of LTFArray with reliability based CMAES learner.

Owner

nils-wisiol Dec 6, 2017

include reference to paper

pypuf/experiments/experiment/reliability_based_cmaes.py Outdated

+                               seed_model, pop_size, step_size_limit, iteration_limit,
+                               ):
+                      """Initialize an Experiment using the Reliability based CMAES Learner for modeling LTF Arrays
+                      :param log_name:        Log name, Prefix of the path or name of the experiment log file

Owner

nils-wisiol Dec 6, 2017

I believe a path prefix is not good here

pypuf/experiments/experiment/reliability_based_cmaes.py Outdated

+                               ):
+                      """Initialize an Experiment using the Reliability based CMAES Learner for modeling LTF Arrays
+                      :param log_name:        Log name, Prefix of the path or name of the experiment log file
+                      :param seed_instance:   PRNG seed used to create LTF array instances

Owner

nils-wisiol Dec 6, 2017

we only init once, i.e. one instance (singular)

pypuf/experiments/experiment/reliability_based_cmaes.py Outdated

+                      :param n:               Length, the number stages within the LTF array
+                      :param noisiness:       Noisiness, the relative scale of noise of instance compared to the scale of weights
+                      :param seed_challenges: PRNG seed used to sample challenges
+                      :param num:             Challenge number, the number of binary inputs for the LTF array

Owner

nils-wisiol Dec 6, 2017

what's this?

pypuf/experiments/experiment/reliability_based_cmaes.py Outdated

+                      :param noisiness:       Noisiness, the relative scale of noise of instance compared to the scale of weights
+                      :param seed_challenges: PRNG seed used to sample challenges
+                      :param num:             Challenge number, the number of binary inputs for the LTF array
+                      :param reps:            Repetitions, the frequency of evaluations of every challenge (part of training_set)

Owner

nils-wisiol Dec 6, 2017

use repetitions instead of frequency, as the latter suggests that this is an ongoing process (which it is not)

pypuf/learner/evolution_strategies/reliability_based_cmaes.py

+                          reliabilities = np.zeros(np.shape(delay_diffs))
+                          indices_of_reliable = np.abs(delay_diffs[:]) > epsilon
+                          reliabilities[indices_of_reliable] = 1
+                          correlation = this.calc_corr(reliabilities, measured_rels)

Owner

nils-wisiol Dec 6, 2017

why not use __class__ here?

pypuf/learner/evolution_strategies/reliability_based_cmaes.py Outdated

+                      return False
+                  @staticmethod
+                  def common_responses(responses):

Owner

nils-wisiol Dec 6, 2017

rename to majority

pypuf/tools.py Outdated

-                      self.challenges = array(list(sample_inputs(instance.n, N)))
-                      self.responses = instance.eval(self.challenges)
+                      self.challenges = array(list(sample_inputs(instance.n, N, random_instance=random_instance)))
+                      if reps is None:

Owner

nils-wisiol Dec 6, 2017

move that to the function definition

pypuf/tools.py Outdated

+                          reps = 1
+                      self.responses = zeros((reps, N))
+                      for i in range(reps):
+                          self.challenges, cs = itertools.tee(self.challenges)

Owner

nils-wisiol Dec 6, 2017

why are use assigning a value to self.challenges here? I think it is unused and later overwritten. (Use _ instead?)

requirements.txt Outdated

@@ @@ -4,3 +4,4 @@ polymath @@
               pylint
               pep8
               pypuf_helper
+              cma

Owner

nils-wisiol Dec 6, 2017 •

edited

Loading

include an exact version here, e.g. 2.3.* (this will update bug fixes but no major changes)

Owner

nils-wisiol commented Jan 17, 2018 •

edited

Loading

current unit test errors, also the tests were running for 12hrs+

Chrism7 force-pushed the cma_hansen branch 2 times, most recently from 65ee569 to 5551579 Compare

January 24, 2018 12:03

Collaborator Author

Chrism7 commented Jan 24, 2018

@nils-wisiol , please review my last commit within this pull request.
Btw: Do you have any idea about how to solve these 3 remaining linter-errors?

nils-wisiol force-pushed the cma_hansen branch from 174dbc3 to 077623b Compare

January 25, 2019 14:44

Owner

nils-wisiol commented Jan 25, 2019 •

edited

Loading

I updated this branch to match up with the updated master branch. It is now, as far as I can tell, running without errors. However, I could not figure out parameters to learn an (relatively easy) instance with n=64 and k=2. For example,

python3 sim_rel_cmaes_attack.py 64 2 1 10000 30 20 20 500 TEST

i.e., n=64, k=2, nosiness 1, 10,000 CPRs, 30 samples per CRP and population size of 20 will stop after the maximum iteration count (500) with an accuracy of around 56%.

seed for instance:      0x157aab5e
seed for challenges:    0xeb576e0a
seed for model:         0x2f8cfd54

Variations of these parameters yield similar results. Could you include a couple of examples that give high accuracy results?

nils-wisiol force-pushed the cma_hansen branch from 077623b to 0004837 Compare

January 25, 2019 15:01

This was referenced Jan 25, 2019

Becker learning #8

Closed

Optimization Based Learning #9

Closed

Chrism7 and others added 14 commits

October 31, 2019 17:30


          try to rebase

176ee19

...

7359cf9


          edit tests in test/test_reliability_based_cmaes.py

e4cec8f


          fix requirements

dee7689


          minor changes

fcb9a7a


          remove usage of mock

7702b10


          delete test.py

a6286d0


          fix TrainingSet in tools.py

fe62db8


          fix initialization of variables

2c9d8a7


          fix all tests

c62d473


          refactor code according to linter

fcafe2a


          fix some linter errors

282991a


          make the linter happy

6098d4c


          bump dependency version

83e9efe

nils-wisiol force-pushed the cma_hansen branch from 0004837 to 83e9efe Compare

October 31, 2019 16:39

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet