Skip to content

MPSFuzz/AI_adversarial_interface

Repository files navigation

Introduction

This code achieves an interface, receiving the POST and GET then scheduling several docker containers where adversarial samples are generated.

Now it implemented to Task 16, docker link and security-enhance modes have been implemented.

Usage

  • setting with ip and port in gunicorn.conf.py, as follows:

    workers = 5

worker_class = "gevent"

bind = "127.0.0.1:5901" // ip and port

  • using run.sh to start the interface:

    chmod a+x run.sh && ./run.sh

Debug mode

//in the url blank:

curl -X POST http://127.0.0.1:5901/adver_gen \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "mission_id=123&test_model=Vgg16&test_weight=weightA&test_seed=seed1&test_method=FGSM&timeout=3600"  \
--noproxy 127.0.0.1     ***if you has set proxy, this option should be added***
curl -X POST http://127.0.0.1:5901/adver_gen_stop \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "mission_id=123" \
--noproxy 127.0.0.1    
curl -X POST http://127.0.0.1:5901/adver_eval \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "mission_id=123" \
--noproxy 127.0.0.1    
curl -X POST http://127.0.0.1:5901/sec_enhance \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "enhance_id=777&test_model=Vgg16&mission_id=12" \
--noproxy 127.0.0.1    

About

There is a interface project to dispatch several docker containers which contain different AI adversarial gen.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors