willer-lu/phishpedia

Official Implementation of "Phishpedia: A Hybrid Deep Learning Based Approach to Visually Identify Phishing Webpages" USENIX'21

Phishpedia A Hybrid Deep Learning Based Approach to Visually Identify Phishing Webpages

Paper • Website • Video • Dataset • Citation

• This is the official implementation of "Phishpedia: A Hybrid Deep Learning Based Approach to Visually Identify Phishing Webpages" USENIX'21 link to paper, link to our website, link to our dataset.

• Existing reference-based phishing detectors:

  • ❌ Lack of interpretability
  • ❌ Lack of generalization performance in the wild
  • ❌ Lack of a large-scale phishing benchmark dataset

• The contributions of our paper:

  • ✅ We propose a phishing identification system Phishpedia, which has high identification accuracy and low runtime overhead, outperforming the relevant state-of-the-art identification approaches.
  • ✅ Our system provides explainable annotations which increase users' confidence in model prediction
  • ✅ We conducted a phishing discovery experiment on emerging domains fed from CertStream and discovered 1,704 real phishing, out of which 1133 are zero-days

Framework

Input: A URL and its screenshot Output: Phish/Benign, Phishing target

• Step 1: Enter Deep Object Detection Model, get predicted logos and inputs (inputs are not used for later prediction, just for explanation)

• Step 2: Enter Deep Siamese Model

  • If Siamese report no target, Return Benign, None
  • Else Siamese report a target, Return Phish, Phishing target

Project structure

- src
    - adv_attack: adversarial attacking scripts
    - detectron2_pedia: training script for object detector
     |_ output
      |_ rcnn_2
        |_ rcnn_bet365.pth 
    - siamese_pedia: inference script for siamese
     |_ siamese_retrain: training script for siamese
     |_ expand_targetlist
         |_ 1&1 Ionos
         |_ ...
     |_ domain_map.pkl
     |_ resnetv2_rgb_new.pth.tar
    - siamese.py: main script for siamese
    - pipeline_eval.py: evaluation script for general experiment

- tele: telegram scripts to vote for phishing 
- phishpedia_config.py: config script for phish-discovery experiment 
- phishpedia_main.py: main script for phish-discovery experiment 

Instructions

Requirements:

• CUDA 11
• Anaconda installed, please refer to the official installation guide: https://docs.anaconda.com/free/anaconda/install/index.html

1. Create a local clone of Phishpedia

git clone https://github.com/lindsey98/Phishpedia.git

2. Setup

cd Phishpedia/
chmod +x ./setup.sh
./setup.sh

If you encounter any problem in downloading the models, you can manually download them from here https://huggingface.co/Kelsey98/Phishpedia. And put them into the corresponding conda environment.

conda activate myenv

Run in Python to test a single website

from phishpedia.phishpedia_main import test
import matplotlib.pyplot as plt
from phishpedia.phishpedia_config import load_config

url = open("phishpedia/datasets/test_sites/accounts.g.cdcde.com/info.txt").read().strip()
screenshot_path = "phishpedia/datasets/test_sites/accounts.g.cdcde.com/shot.png"
ELE_MODEL, SIAMESE_THRE, SIAMESE_MODEL, LOGO_FEATS, LOGO_FILES, DOMAIN_MAP_PATH = load_config(None)

phish_category, pred_target, plotvis, siamese_conf, pred_boxes = test(url=url, screenshot_path=screenshot_path,
                                                                       ELE_MODEL=ELE_MODEL,
                                                                       SIAMESE_THRE=SIAMESE_THRE,
                                                                       SIAMESE_MODEL=SIAMESE_MODEL,
                                                                       LOGO_FEATS=LOGO_FEATS,
                                                                       LOGO_FILES=LOGO_FILES,
                                                                       DOMAIN_MAP_PATH=DOMAIN_MAP_PATH
                                                                      )

print('Phishing (1) or Benign (0) ?', phish_category)
print('What is its targeted brand if it is a phishing ?', pred_target)
print('What is the siamese matching confidence ?', siamese_conf)
print('Where is the predicted logo (in [x_min, y_min, x_max, y_max])?', pred_boxes)
plt.imshow(plotvis[:, :, ::-1])
plt.title("Predicted screenshot with annotations")
plt.show()

Or run in bash

python run.py --folder <folder you want to test e.g. phishpedia/datasets/test_sites> --results <where you want to save the results e.g. test.txt> 

Miscellaneous

• In our paper, we also implement several phishing detection and identification baselines, see here
• The logo targetlist described in our paper includes 181 brands, we have further expanded the targetlist to include 277 brands in this code repository
• For the phish discovery experiment, we obtain feed from Certstream phish_catcher, we lower the score threshold to be 40 to process more suspicious websites, readers can refer to their repo for details
• We use Scrapy for website crawling Repo here

Citation

If you find our work useful in your research, please consider citing our paper by:

@inproceedings{lin2021phishpedia,
  title={Phishpedia: A Hybrid Deep Learning Based Approach to Visually Identify Phishing Webpages},
  author={Lin, Yun and Liu, Ruofan and Divakaran, Dinil Mon and Ng, Jun Yang and Chan, Qing Zhou and Lu, Yiwen and Si, Yuxuan and Zhang, Fan and Dong, Jin Song},
  booktitle={30th $\{$USENIX$\}$ Security Symposium ($\{$USENIX$\}$ Security 21)},
  year={2021}
}

Contacts

If you have any issues running our code, you can raise an issue or send an email to [email protected], [email protected], and [email protected]