Keystroke Pattern Authentication Using Outlier Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing keystroke pattern-based user authentication systems face challenges in accurately identifying users when using arbitrary text input and are affected by outliers, leading to reduced performance and vulnerability to impersonation, especially when trying to verify users on arbitrary or 'free' text inputs.
Innovation Solution
The method involves creating a training data set of keystroke patterns, removing outliers, and using either the Competition between Naïve Bayes Models (CNBM) or Similarity-based user identification methods to compare test data against authorized user profiles, allowing for identification on arbitrary text and enhancing outlier detection and removal processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If keystroke patterns are used for user authentication on arbitrary text inputs, then the system becomes more versatile and less intrusive, but identification accuracy decreases due to variability in typing behavior and presence of outliers
Solution Approach 1:
The system performs preliminary outlier detection and removal from training data before creating user profiles. This preliminary action ensures that the profiles are built from clean, representative data, improving identification accuracy while maintaining the ability to handle arbitrary text inputs.
Solution Approach 2:
The system extracts and removes outliers from the keystroke pattern data set before profile creation. By taking out these abnormal data points that don't represent typical user behavior, the system improves measurement precision without sacrificing versatility in handling different text inputs.
2Measurement precision
If fixed text strings are used for creating user profiles, then identification accuracy improves, but the system becomes vulnerable to impersonation and less secure
Solution Approach 1:
The system transitions from static fixed text strings to dynamic arbitrary text inputs for authentication. By analyzing keystroke patterns on variable text inputs, the system maintains high identification accuracy while improving security, as imposters cannot replicate the dynamic typing behavior across different texts.
Solution Approach 2:
The system changes the parameter of text input from fixed to arbitrary/variable. This parameter change allows the system to maintain measurement precision through statistical analysis of keystroke patterns while significantly improving reliability by making impersonation much more difficult.
3Quantity of substance
If outliers are retained in the training data set, then more data is available for profile creation, but identification accuracy decreases due to noisy or unrepresentative patterns
Solution Approach 1:
The system converts the harmful effect of outliers (noisy data reducing accuracy) into a benefit by using outlier detection as a feature. The outlier detection process itself becomes a valuable step that filters data quality, ensuring that only representative keystroke patterns are used for profile creation, thereby improving identification accuracy.
Data Source
AI summary
A method of identifying a user as an authorized user from free test text typed by that user into an input device. From the received test text, features associated with the typed text are extracted, such as timing data associated with alphanumeric letter pairs. These extracted features are compared to previously stored series of authorized user profiles, where the authorized user profiles were generated from a trial typing sample of alphanumeric data from each associated authorized user. The comparison identifies one of the authorized users with the user, and a score is derived to measure the strength of the comparison. If the score exceeds a threshold level, the user is identified as that authorized user.