Gesture Fingerprinting for Bot Detection in Touchscreen Inputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital advertising campaigns face significant fraudulent clicks and views due to automated bots mimicking human interactions, leading to inflated ad revenue and invalid impression statistics, with existing technologies lacking reliable methods to distinguish between human and bot inputs.
Innovation Solution
A computer-implemented method using gesture fingerprinting technology that captures sensor data from touchscreen inputs, compares variance criteria to determine whether the input is from a human user or a programmed device, and stores the determination as analytics data, employing a system with processors, memories, and variability profiles to differentiate between human and bot interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If applications count all inputs as valid user interactions, then ad revenue metrics are inflated, but the accuracy of analytics deteriorates
Solution Approach 1:
The patent changes the parameter being measured from simple input detection to variance analysis of sensor data. By analyzing the variance characteristics of sensor readings over time, the system can distinguish between human inputs (which exhibit natural variance patterns) and bot inputs (which exhibit different variance patterns), thereby improving the reliability of ad analytics without losing valid impression data
Solution Approach 2:
The patent replaces the mechanical/manual method of distinguishing human vs. bot inputs with an automated sensor-based detection system. The system uses sensors to capture input characteristics and automatically analyzes variance patterns to identify bot behavior, eliminating the need for manual verification and providing continuous automated fraud detection
2Measurement precision
If variance detection thresholds are set to be highly sensitive, then bot detection accuracy improves, but false positives increase reducing valid human inputs
Solution Approach 1:
The patent applies partial action by using multiple variance criteria instead of a single threshold. The system evaluates multiple aspects of input variance (temporal patterns, spatial patterns, sensor-specific patterns) and requires satisfaction of multiple criteria to classify an input as bot-generated. This partial approach allows the system to maintain high detection accuracy while reducing false positives by not relying on any single overly-sensitive threshold
3Difficulty of detecting and measuring
If comprehensive sensor data is collected for all inputs, then detection capability is enhanced, but system complexity and processing overhead increase
Solution Approach 1:
The patent extracts only the essential variance characteristics from comprehensive sensor data rather than processing all raw sensor information. The system identifies and extracts key variance patterns (temporal variance, spatial variance, pressure variance) that are most indicative of human vs. bot behavior, discarding redundant data. This extraction approach maintains strong detection capability while reducing system complexity and processing overhead
Data Source
AI summary
Various implementations related to human versus bot detection using gesture fingerprinting are described. In one such implementation, a computer-implemented method includes receiving an input associated with an application presented on a display device, capturing sensor data associated with the input, detecting a variance in the sensor data, comparing the variance with variance criteria, determining whether the input is provided by a human user or mimicked by a programmed device based on the comparison of the variance with the variance criteria, and storing the determination as analytics data in association with the application.