Dynamic Client Device Classification via Event Timing Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying and classifying client devices accessing content from a server are inadequate in speed and accuracy, particularly in distinguishing between legitimate users and malicious bots, as they often require fixed time periods and do not account for the variability in event recording times and orders, leading to potential misclassification and delayed detection of bot activities.
Innovation Solution
A computer-implemented method that monitors and stores event vectors associated with client device interactions, determining a classification time based on the evidence vector to establish a predetermined level of confidence in classifying the type of agent controlling the client device, allowing for real-time and accurate differentiation between legitimate users and bots by modeling the expected timing and order of events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If fixed time periods are used for classifying client devices, then the classification process is simple to implement, but the accuracy and speed of bot detection deteriorates due to variability in event recording times and orders
Solution Approach 1:
The patent implements dynamic classification time windows that adapt to the actual timing and ordering of events observed in client device interactions. Instead of using fixed time periods, the system calculates classification time windows based on the temporal patterns and sequences of events specific to each client device, allowing the classification process to adjust dynamically to varying event recording times and orders, thereby maintaining high accuracy while remaining computationally efficient
Solution Approach 2:
The system changes the parameter of classification time from a fixed value to a dynamic value determined by event timing patterns. By analyzing the temporal characteristics of events (such as page load times, interaction intervals, and event sequences) and adjusting the classification time window accordingly, the system achieves accurate bot detection without requiring overly complex implementation
2Ease of manufacture
If fixed time periods are used for classifying client devices, then the implementation is straightforward, but the speed of bot detection deteriorates due to delayed classification
Solution Approach 1:
The system employs dynamic time windows that are optimized for each classification case, allowing rapid classification when event patterns are clear and consistent. By adapting the classification time window to the actual event timing, the system can quickly identify bots without waiting for fixed time periods to elapse, significantly improving detection speed while maintaining implementation simplicity through algorithmic efficiency
3Device complexity
If event timing and order variability is not accounted for, then the classification process is simpler, but misclassification increases leading to unreliable bot identification
Solution Approach 1:
The patent applies preliminary actions by pre-establishing the expected timing and ordering patterns of events for legitimate client devices. The system creates reference profiles of normal event sequences and timing characteristics before classification occurs. During classification, actual event patterns are compared against these pre-established references, allowing the system to account for timing and order variability without requiring complex real-time analysis, thus maintaining relatively simple processing while achieving reliable bot identification
Solution Approach 2:
The system incorporates feedback mechanisms that analyze the timing and ordering of events and adjust classification decisions accordingly. By providing feedback on whether observed event patterns conform to expected temporal sequences, the system can reliably distinguish bots from legitimate devices even when some variability exists, improving reliability without proportionally increasing complexity
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A computer-implemented method for classifying an agent controlling a client device requesting content from a server is disclosed. The method comprises monitoring, by a classification server (30), a plurality of events associated with an interaction made by a client device (10) with a server (20), storing, by the classification server (30) in a memory, a record of the monitored events in an order and timing in which the events occurred, comparing, by the classification server (30), the record of the monitored events against probabilistic models of expected inter-event times, determining, by the classification server (30) based on the comparison, a classification time representing a time at which a probability that the classification server can classify an agent controlling the client device with a predetermined level of confidence, and classifying, by the classification server (30), the agent controlling the client device as legitimate or illegitimate based on the determined classification time and the order and timing of events in the record.