AMR_Adversarial_Attacks

This is a repository includes the source codes of paper "Security Concerns of Adversarial Attacks in Machine Learning Based Automatic Modulation Recognition Models for Next Generation Networks".

Installation & Usage

• Run AMR for SISO Dataset:
  1. Download the SISO dataset (See RML2016.10a) and save it under AMR_SISO folder. RML2016.10a has been used in most research as a benchmark dataset. It contains 220,000 signal samples and each sample is associated with one specific modulation (11 modulations in total) at a particular Signal-to-Noise Ratio (SNR range: -20dB:2:18dB). Each sample input is a vector of size 256, which corresponds to 128 in-phase and 128 quadrature components.

2. Change the dataset filename in line 55 in rmldataset2016.py.

• Run AMR for SISO NEW Dataset:
  1. Download the SISO dataset (See RML2016.10b) and save it under AMR_SISO folder.

2. Change the dataset filename in line 55 in rmldataset2016.py.