search

TanGentleman / KeyMaster

A comprehensive toolkit for python classes and easy to use scripts for logging, analyzing, and simulating keystrokes.
MIT License
1 stars 0 forks source link

Use non-arbitrary values for keystroke generation algorithm #6

Open TanGentleman opened 8 months ago

TanGentleman commented 8 months ago

Current normalization is 60ms with 15ms standard deviation. This works, but should be restructured to plan for future compatibility with datasets and sophisticated prediction.

TanGentleman commented 7 months ago

This has been on the back burner a while, but I'll be working on this alongside polishing the client interface.

TanGentleman commented 7 months ago

Progress update, here's the fancy approach we're taking to this problem: Relevant paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10474054/ Dataset preview: https://zenodo.org/records/7886743/preview/free-text.csv?include_deleted=0

I'm going to be needing a new class to handle this dataset, and it's gonna have some cool methods! Assuming I can reasonably make sense of it, clean it (transparently), and extract some meaningful metrics, I still have some issues ahead. I have to convert my findings into a meaningful predictive algorithm, and then adjust my implementation to actually follow through with it. There's a few unknowns here, but if it works, then instead of using keyboard.tap in my implementation, I can increase the precision between the press and release. I'll update as I make meaningful headway.

TanGentleman commented 7 months ago

A great resource that treads more in the realm of cybersecurity, but contains valuable insights:

  1. https://www.cs.cmu.edu/~keystroke/

And an even more comprehensive ones that are really going to be useful:

  1. https://www.researchgate.net/publication/228103101_Web-Based_Benchmark_for_Keystroke_Dynamics_Biometric_Systems_AStatistical_Analysis

  2. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8782503/