Detecting Usage Protocol Violations in Deep Learning Applications

<p>This repo contains a replication package of the "Detecting Usage Protocol Violations in Deep Learning Applications" paper.</p><p>In the paper, we use two benchmarks: &nbsp;</p><ul><li>The first benchmark we used from NeuraLint can be found under the director name Benchmark1/SOSamples</li><li>The second benchmark we used from Humbatova et al. can be found under the director name Benchmark2/SOSamples</li></ul><p>To reproduce the results in the paper: &nbsp;</p><ul><li>Download the NeuralStateAnalysis Zip file.</li><li>Extract the file and go to the NeuralStateAnlaysis directory.</li><li>( Optional ) Install the requirements by running 'Pip install requirements.txt.' N.B: The requirements.txt file is already in this repo.<ul><li>To run NeuralState on Benchmark1:<ul><li>Go to Benchmark1/SOSamples directory,</li><li>Open any of the programs you want to run,</li><li>Set the path: Path-to-folder/NeuralStateAnalysis/</li><li>Then, do 'python program_id.' Since the 'NeuralStateAnalysis(model).debug()' call is already present in all programs, you'll be able to reproduce results.</li></ul></li><li>To run NeuralState on Benchmark2:<ul><li>Go to Benchmark2/SOSamples directory,</li><li>Open any of the programs you want to run,</li><li>Set the path: Path-to-folder/NeuralStateAnalysis/</li><li>Then, do 'python program_id.' Since the 'NeuralStateAnalysis(model).debug()' call is already present in all programs, you'll be able to reproduce results.</li></ul></li></ul></li></ul>