Keystroke Biometric Authentication Using Bigram Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional keystroke authentication systems are inefficient and require high performance for accurate verification, especially in free-text input scenarios, and are not suitable for quick and reliable authentication in conversational interfaces.
Innovation Solution
A user authentication method using Gaussian Mixture Models (GMM) with Universal Background Models (UBM) to analyze keystroke dynamics by separating key presses into bigrams, collecting timing information, extracting feature vectors, and estimating GMM user models for real-time authentication, allowing for low-power device implementation and language-independent authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional keystroke authentication systems are used, then user verification can be performed, but the systems require high performance and are time-consuming
Solution Approach 1:
The patent segments the keystroke authentication process into distinct phases: enrollment phase where user-specific parameters are captured and stored, and verification phase where these parameters are compared against new keystroke data. This segmentation allows the system to pre-process and store user biometric patterns during enrollment, reducing the computational burden during actual verification operations. The authentication system divides the feature extraction into character-level and word-level analyses, enabling efficient comparison without requiring high-performance processing during verification.
Solution Approach 2:
The system performs preliminary action by capturing and storing user-specific keystroke parameters during an enrollment phase before actual authentication is needed. During enrollment, the system collects timing information, key press durations, and typing patterns, then processes this data to create user-specific profiles. This preliminary processing stores the computational work in accessible form, so that during verification, the system only needs to compare new keystroke data against pre-computed profiles rather than performing complex analysis in real-time.
2Reliability
If neural networks are used for keystroke verification, then authentication can be performed, but the algorithm is heavyweight and requires imposters' samples for training
Solution Approach 1:
The patent extracts and utilizes specific keystroke dynamics parameters such as timing information between key presses, key press durations, and typing rhythms. Rather than using complex neural networks that require extensive training data including imposter samples, the system extracts meaningful biometric features directly from legitimate user keystroke patterns. This extraction approach focuses on capturing essential user-specific characteristics without the overhead of training complex models, achieving verification accuracy through targeted feature measurement rather than brute-force pattern recognition.
Solution Approach 2:
The system changes parameters by focusing on timing-based metrics such as the time interval between key presses, duration of key press events, and rhythmic patterns of typing. These parameter changes transform the authentication approach from content-based verification (what is typed) to temporal-based verification (how it is typed). By measuring and comparing timing parameters across different keystroke sequences, the system creates a robust biometric profile that is resistant to spoofing while requiring minimal computational resources compared to neural network approaches.
3Reliability
If Support Vector Machine approach is used, then keystroke authentication can be performed, but more powerful software is required and accuracy is affected
Solution Approach 1:
The patent implements self-service by using user-specific keystroke parameters captured during normal typing activities to automatically create verification profiles. During the enrollment phase, the system collects timing data and typing patterns from legitimate users, then automatically processes this data to generate user-specific authentication profiles without requiring manual configuration or complex software intervention. The system serves itself by utilizing the natural keystroke data generated during normal user interaction to build and maintain verification models, eliminating the need for powerful external processing software.
Data Source
AI summary
A method for user authentication based on keystroke dynamics is provided. The user authentication method includes receiving a keystroke input implemented by a user; separating a sequence of pressed keys into a sequence of bigrams having bigram names simultaneously with the user typing free text; collecting a timing information for each bigram of the sequence of bigrams; extracting a feature vector for each bigram based on the timing information; separating feature vectors into subsets according to the bigram names; estimating a GMM user model using subsets of feature vectors for each bigram; providing real time user authentication using the estimated GMM user model for each bigram and bigram features from current real time user keystroke input. The corresponding system is also provided. The GMM based analysis of the keystroke data separated by bigrams provides strong authentication using free text input, while user additional actions (to be verified) are kept at a minimum. The present invention allows to drastically improve accuracy of user authentication with low performance requirements that allows to implement authentication software for low-power mobile devices.