Keystroke Pattern Analysis Using Time-Based Vector Representations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current text-based identity verification systems are weak and prone to information leakage, and keystroke pattern analysis methods face inefficiencies in feature extraction and classification accuracy due to reliance on expert knowledge and difficulties in characterizing time-ordered keystroke information.
Innovation Solution
A method that encodes keystroke information into vectorized representations, incorporating time information to create time-based vectorized representations, which are then used to extract keystroke behavior features for user verification through a machine learning model, improving feature extraction and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If text-based identity verification is used, then the authentication process is simple and straightforward, but the security mechanism is weak and prone to information leakage
Solution Approach 1:
The patent combines text-based verification with keystroke dynamics analysis into a unified authentication system. The system processes both the content of typed characters and the temporal patterns of keystrokes simultaneously, merging two verification approaches to achieve both simplicity and enhanced security
2Measurement precision
If keystroke pattern analysis is performed using traditional methods, then user behavior is monitored, but feature extraction efficiency is low and classification accuracy is insufficient due to reliance on expert knowledge
Solution Approach 1:
The patent replaces manual expert-defined feature extraction with an automated machine learning model. The system uses supervised learning algorithms to automatically learn and extract keystroke behavior features from raw keystroke data, eliminating the need for manual feature engineering while improving both accuracy and efficiency
Solution Approach 2:
The patent transforms raw keystroke data into a structured feature space by extracting multiple temporal and behavioral parameters (key press duration, key release duration, pause times, typing speed). This parameter transformation enables the machine learning model to process and classify keystroke patterns effectively
3Reliability
If keystroke information is collected and analyzed, then user identity verification is achieved, but the complexity of processing time-ordered keystroke information increases
Solution Approach 1:
The patent segments the keystroke authentication process into distinct modules: data collection, feature extraction, and classification. Each module handles a specific aspect of the processing pipeline, reducing overall system complexity while maintaining verification reliability
Data Source
AI summary
Embodiments of the present disclosure provide a method, a device, and a program product for keystroke pattern analysis. The method includes: acquiring keystroke information of a user on an electronic device, wherein the keystroke information indicates a sequence of characters that are typed sequentially and time information related to the typing of corresponding characters in the sequence of characters; encoding corresponding characters in the sequence of characters respectively into vectorized representations to obtain a sequence of vectorized representations, wherein different characters are encoded into different vectorized representations; superimposing the time information related to the typing of corresponding characters in the sequence of characters respectively to corresponding vectorized representations in the sequence of vectorized representations to obtain a sequence of time-based vectorized representations; and verifying a keystroke pattern of the user by extracting keystroke behavior features from the sequence of time-based vectorized representations.