Skip to content

Humanize Behaviour v4 Submission Guide (Active after June 9th 2025 14:00 UTC)

Overview

Humanize Behaviour v4 tests bot scripts' ability to mimic human interaction with a web UI form with dynamic movements accross different sessions. It evaluates how well a bot navigates UI elements, interacts with form fields, and submits data without being caught by the bot detection system, based on mouse movement, keyboard interaction analysis. and mouse movement similarities over sessions

Miners must demonstrate precise, human-like interactions through their bot scripts when completing the form. And also they have to keep in mind their bots should not make same mouse movement trajectory in each iteration

Example Code and Submission Instructions

Example codes for the Humanize Behaviour v4 can be found in the redteam_core/miner/commits/humanize_behaviour_v4/ directory.

Technical Requirements

  • Python 3.10
  • Ubuntu 24.04
  • Docker container: selenium/standalone-chrome:4.28.1

Core Requirements

  1. Use our template from redteam_core/miner/commits/humanize_behaviour_v4/src/bot/bot.py
  2. Keep the run_bot() function signature unchanged
  3. Your bot must:
  4. Work with the provided Selenium driver
  5. Input text into designated fields
  6. Scroll down to the end of the page
  7. Click the designated button to end the session
  8. Submit the form without errors

Key Guidelines

  • Driver Usage: Stick to the provided Selenium driver to ensure proper evaluation
  • Action Sequence: Follow the provided ACTIONS LIST order. Clicking it at the start or in the middle will prematurely submit data and result in a zero score due to invalid action flow.
  • Click Behavior:
    • Only click at specified locations
    • Additional clicks for input fields and submit buttons are allowed
    • Wrong click order will result in form submission failure.
  • Text Input:
    • Locate fields by their id
    • Use text from the ACTIONS LIST
    • Maintain the specified input order
  • Page Interaction:
    • After completing form inputs, scroll down to the end of the page
    • Click the designated button to properly end the session
    • This simulates natural user flow and ensures full-page interaction capability
  • Cross-Session Behavior Variation:
    • Your bot must demonstrate different movement patterns across multiple evaluation sessions
    • While click targets remain the same, mouse trajectories should vary naturally between sessions
    • Similar movement patterns across sessions will be flagged as bot-like behavior
    • Implement randomization strategies that mimic natural human variability
  • Technical Setup:
    • Enable headless mode
    • List dependencies in pip_requirements. See the limitations for dependencies below.
    • Use amd64 architecture (ARM64 at your own risk)
    • If your script requires system-level dependencies, add them to system_deps field as ubuntu packages (concatenate with space if multiple packages are needed "package1 package2") in /solve endpoint reponse.
  • Limitations
    • Your script must not exceed 2,000 lines. If it does, it will be considered invalid, and you will receive a score of zero.
    • Your dependencies must be older than January 1, 2025. Any package released on or after this date will not be accepted, and your script will not be processed.

Evaluation Criteria

Your bot will be scored on these human-like behaviors:

  • Mouse Movement Velocity Variation
  • Mouse Movement Speed
  • Mouse Movement Velocity Profiles between clicks
  • Mouse Movement Granularity (average pixel per movement)
  • Mouse Movement Count within the session
  • Mouse Movement Trajectory Linearity
  • Keypress Behavior Pattern (typing speed and variations)
  • Scroll Path Linearity(Scroll should be linear in single coordinate)

Plagiarism Check

We maintain strict originality standards:

  • All submissions are compared against other miners' script
  • 100% similarity = zero score
  • Similarity above 60% will result in proportional score penalties based on the detected similarity percentage.
  • Note: Comparisons are only made against other miners submissions, not your own previous Humanize Behaviour v3 entries.

Testing Solutions

You can see testing guidelines here in humanize_behaviour_v4/docs/testing_manual.md. After testing you can process with Submission guide documentation or Submission guide

Submission Guide

Follow 1~6 steps to submit your script.

  1. Navigate to the Humanize Behaviour v4 Commit Directory

    cd redteam_core/miner/commits/humanize_behaviour_v4

  2. Build the Docker Image

    To build the Docker image for the Humanize Behaviour v4 submission, run:

    docker build -t my_hub/humanize_behaviour_v4-miner:0.0.1 .

# For MacOS (Apple Silicon) to build AMD64:
DOCKER_BUILDKIT=1 docker build --platform linux/amd64 -t myhub/humanize_behaviour_v4-miner:0.0.1 .

  3. Log in to Docker

    Log in to your Docker Hub account using the following command:

    docker login

    Enter your Docker Hub credentials when prompted.

  4. Push the Docker Image

    Push the tagged image to your Docker Hub repository:

    docker push myhub/humanize_behaviour_v4:0.0.1

  5. Retrieve the SHA256 Digest

    After pushing the image, retrieve the digest by running:

    docker inspect --format='{{index .RepoDigests 0}}' myhub/humanize_behaviour_v4:0.0.1

  6. Update active_commit.yaml

    Finally, go to the neurons/miner/active_commit.yaml file and update it with the new image tag:

    - humanize_behaviour_v4---myhub/humanize_behaviour_v4@<sha256:digest>

📑 References