Covert Channel

Documentation

Structure

• src/: source code (Uses the top-level CMake build system)
  • experiments/: standalone experiments
    • receiver: receiver application
      • main.cpp receiver application entry point
    • timing-consistency: timing consistency benchmark
      • main.cpp timing consistency benchmark entry point
    • transmitter/: transmitter application
      • main.cpp transmitter application entry point
  • lib/: library source code
    • contention_generator.cpp: contention generator implementation
    • contention_generator.hpp: contention generator definition
    • precise_sleep.cpp: precise sleep implementation
    • precise_sleep.hpp: precise sleep definition
• scripts/: Experiments and Data analysis (runs transmitter/receiver for experiments and analyzes the traces)
  • buffer_size_experiment.py Changes the buffer size on both transmitter and receiver and runs combinations of experiments
  • transmit_rate_experiments.py Changes the frequency of data rate and runs experiments
  • analysis.py Calculates the accuracy of experiments
  • print_bandwidth.py Prints only bandwidths
  • emc_values.py Calculates average/max EMC values

Setup

1. Build everything:

cmake -S . -B build
cmake --build build --parallel

2. Copy and compile the GPU code. We basically modified the bandwidthTest.cu sample from CUDA samples and used it as GPU receiver. Note (1st command): Paths may not match. In this case, just copy samples in cuda under ~/cuda_samples Note (2nd command): Paths may not match. In this case, just the bandwidthTest.cu in our repo to cuda samples (replace it)

cp -r /usr/local/cuda/samples ~/cuda_samples
cp src/GPU_samples/bandwidthTest.cu ~/cuda_samples/1_Utilities/bandwidthTest/bandwidthTest.cu
cd ~/cuda_samples/1_Utilities/bandwidthTest
make

3. Run experiments for cpu (tranmistter) to gpu (receiver):

This experiment evaluates receiver buffer size. (sensitivity analysis of receiver). More parameters can be modified to perform more experiments, some of which are left under comments as suggestions.

mkdir -p logs/cpu_to_gpu_receiver_buffer_size
python3 scripts/cpu_gpu.py --log_dir=logs/cpu_to_gpu_receiver_buffer_size
python3 scripts/analysis_microsecond.py --log_dir=logs/cpu_to_gpu_receiver_buffer_size --calculateAccuracy=History --threshold=0.05 --mergeFile=True