Choose Prediction Method

Train New Model

Trigger a new training pipeline to update the model

Batch Prediction

Upload a CSV file containing URL features for bulk analysis

Drag & Drop CSV file here

or click to browse
filename.csv

Manual Input

Enter URL features manually to check for phishing

having_IP_Address -1 / 0 / 1
URL_Length -1 / 0 / 1
SSLfinal_State -1 / 0 / 1
+ 27 more features
Open Manual Input

Required Features (30 Columns)

URL Structure

  • having_IP_Address
  • URL_Length
  • Shortining_Service
  • having_At_Symbol
  • double_slash_redirecting
  • Prefix_Suffix
  • having_Sub_Domain
  • Abnormal_URL

Security

  • SSLfinal_State
  • HTTPS_token
  • port
  • Favicon
  • Statistical_report

Page Content

  • Request_URL
  • URL_of_Anchor
  • Links_in_tags
  • SFH
  • Submitting_to_email

Page Behavior

  • Redirect
  • on_mouseover
  • RightClick
  • popUpWidnow
  • Iframe

Domain Info

  • Domain_registeration_length
  • age_of_domain
  • DNSRecord
  • web_traffic
  • Page_Rank
  • Google_Index
  • Links_pointing_to_page
Value Range: All features should be integers with values: -1 (Phishing indicator), 0 (Suspicious), or 1 (Legitimate indicator)

How It Works

1

Data Ingestion

Raw data pulled from MongoDB and split into train/test sets

2

Validation

Schema validation & KS drift detection

3

Transformation

KNN Imputation & target mapping

4

Model Training

5 models evaluated, best selected by F1 score

5

Deployment

Model served via FastAPI on Hugging Face