Skip to content

Latest commit

 

History

History
23 lines (16 loc) · 1.22 KB

README.md

File metadata and controls

23 lines (16 loc) · 1.22 KB
Raw

Code for the Targeted Data Poisoning Attack

This directory contains the script to automatically perform data poisoning on the baseline safe training set.

The README file is written based on our setup experience on Ubuntu 18.04.3 LTS.

To run the code, use python data_poisoning_attack.py [VULN_CATEG] [N].

The script takes 2 arguments:

  1. [VULN_CATEG] Vulnerability category ("ICI", "DPI", or "TPI")
  2. [N], Number of samples to poison (5, 10, 15, ..., 40)

Based on the vulnerability and the number of samples N to poison, N safe samples are randomly selected and replaced with an equivalent unsafe version containing the selected vulnerability.

The final poisoned training is stored in the same directory, divided as follows:

  • PoisonPy_train.in, containing the intents of the training set;
  • PoisonPy_train.out, containing the code snippets of the training set;
  • PoisonPy_dev.in, containing the intents of the validation set;
  • PoisonPy_dev.out, containing the code snippets of the validation set.

The test set is stored in the Dataset/Testset directory, divided as follows:

  • PoisonPy_test.in, containing the intents of the test set;
  • PoisonPy_test.out, containing the code snippets of the test set.