Phishing attacks are regarded as one of the most effective and low effort tools hackers rely on to compromise cybersystems or leak sensitive data. Pharec aims to bring innovative (autonomous) cybersecurity products that improve safety of our daily digital interactions (e.g. make browsing less risky).

Pharec created by a group of highly technical individuals (computer scientists, engineers and researchers) with expertise and skills in both cybersecurity and deep learning. We train your machine to read, see and analyze like a cybersecurity expert beside you.

Deep Learning Technology

End to end deep learning with both image and text data to train a highly accurate phishing detection model. The model has similar visibility (actually greater visibility) to a human and if trained correctly can learn skills of a cybersecurity expert. It is as if you have a cybersecurity expert watching over (helping) you at all times.

Pharec Technology

Features

  • Real time detection of phishing pages.
  • Easy to setup and use (autonomous).
  • Improved detection accuracy compared to other offerings.
  • Continuous improvements (higher quality data and
  • model) adapting to the current landscape of threats.
  • Support for Arabic content (phishing) by
  • including Arabic pages in training.

Prototype demo (MVP)

Check out the demo submitted to Thakaa Cybersecurity Challenge 2021 in youtube. Also a video testing the model here.

Github Repos

We are still in early phases of development (under active development). Please check out the following repos and reach out on social media (found bellow) or through Github (by creating an issue).

  • WebExtension repo
    • Experimental code (not stable) of the phishing image classification model used in a browser extension as submitted to the challenge. Currently works on firefox and can be converted to work on other browsers.
  • Pharec model repo
    • Training notebooks of the deep image CNN classification (prototype) model (training notebooks) of the phishing site classifier. Also includes starter utils for dataset collection