Aishwarya
https://medium.com/personaldata-io/coursera-and-keystroke-biometrics-550762f2f61b ()
- Online platforms can use keystroke biometrics while conducting tests. Coursera.
- Dwell time on each key and flight time seem to be the main parameters.
- There seems to be software to record this.
- Error pattern can also be used as one of the feature, don't know how to integrate though.
- Also how to decide the hyper-parameters.
- General information on Keystroke dynamics: http://www.biometric-solutions.com/keystroke-dynamics.html
- Timing features that can be considered: https://www.researchgate.net/figure/Keystroke-metrics-latency-interval-dwell-time-and-flight-time-Generally-typing_fig1_221247068
- Paper for keystroke dynamics using deep learning. Feature learning with deep neural networks for keystroke dynamics paper
- Companies involved
- TypingDNA
- ID control
- BehavioSec
- Keystroke pattern of individuals by lot of external factors. So is it better if we continuously keep learning the pattern. Eg. for moocs, track the pattern even when the course goes on. (Still has some loop holes though.)
- 2 more metrics that can be used:
- Latency
-
Interval
- dataset and reference, Carnegie Mellon University, https://www.cs.cmu.edu/~keystroke/
- Learn on the go, like in typingDNA.
- Places where it can be implemented, users don't want to type long phrases for small work or any work for that matter.
- Paper: Carnegie Mellon University, [https://www.cs.cmu.edu/~maxion/pubs/KillourhyMaxion09.pdf] (https://www.cs.cmu.edu/~maxion/pubs/KillourhyMaxion09.pdf)
- Github page with code in python: https://github.com/njanakiev/keystroke-biometrics/blob/master/keystroke-biometrics.ipynb
- Dataset for keystroke(some competition): https://sites.google.com/site/btas16kboc/home