Diurnal Behavior Analysis for Bot Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online systems face challenges in distinguishing between human users and non-human users, such as bots, which can negatively impact user experience and system integrity by automating interactions and compromising data security.
Innovation Solution
A machine-learning based system that differentiates between human and bot users by analyzing diurnal behavior patterns in user activity data, transforming time-series data into the frequency domain, and using machine-learning models to classify users as human or non-human based on observed behavior patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional user verification methods are used, then system operation is simple, but the system cannot effectively distinguish between human users and bots
Solution Approach 1:
The patent replaces traditional mechanical verification methods (captchas, manual review) with a machine learning-based automated classification system. The system uses diurnal behavior pattern analysis and time-series data transformation to automatically distinguish human users from bots, substituting manual or rule-based mechanisms with intelligent algorithms that provide higher classification accuracy without requiring user intervention.
Solution Approach 2:
The patent introduces an intermediary classification system that acts as a mediator between user activity data and system responses. This intermediary layer analyzes behavioral patterns, transforms time-series data into frequency domain representations, and provides classification results to the system, enabling accurate bot detection while maintaining system architecture simplicity through a dedicated intermediate processing layer.
2Productivity
If manual bot detection methods are used, then system complexity is low, but productivity in combating bot activities is insufficient
Solution Approach 1:
The patent implements a self-service detection system where the machine learning model automatically analyzes user behavior patterns and classifies users as human or bot without requiring manual intervention. The system self-trains on collected data, continuously improving its detection capabilities while autonomously combating bot activities, thereby achieving high productivity in bot detection while managing complexity through automation.
Solution Approach 2:
The patent performs preliminary classification of users based on diurnal behavior patterns before bot activities can cause significant harm. By continuously monitoring and pre-classifying user behavior, the system identifies potential bots early in their operation, enabling proactive measures to be taken against bot activities before they can compromise system integrity or steal data.
3Measurement precision
If comprehensive user behavior monitoring is implemented, then measurement precision of user classification improves, but loss of user privacy increases
Solution Approach 1:
The patent extracts only the essential diurnal behavior pattern features needed for classification while leaving out unnecessary personal information. The system focuses on extracting temporal activity patterns (when users are active, their rhythm) rather than collecting comprehensive personal data, thereby achieving accurate bot detection while minimizing privacy intrusion by taking out only the necessary behavioral characteristics.
Solution Approach 2:
The patent applies different levels of monitoring intensity to different aspects of user behavior. Instead of uniformly monitoring all user activities with the same depth, the system focuses local analytical resources on diurnal patterns and time-series behavior characteristics that are most indicative of bot activity, while applying lighter monitoring to other aspects, thereby improving classification precision where needed while preserving privacy in other areas.
Data Source
AI summary
In one embodiment, a system is configured to identify, based on predetermined criteria, a first set of users of an online system who belong to a population segment. The system may monitor activities performed by the first set of users on the online system over a predetermined period of time and store the monitored activities as time-series data. A feature set associated with the first set of users may be generated by transforming the time-series data into a frequency domain. The system may train a machine-learning model using the feature set and other feature sets to determine whether activities associated with a given set of users exhibit diurnal behavior pattern. Using the trained machine-learning model, the system may determine whether activities performed by a second set of users on the online system exhibit diurnal behavior pattern.