Human-Behavior Traffic Generation for Encrypted Network Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating high-quality training data for machine-learning modules to classify network traffic is challenging due to enhanced encryption and obfuscation, with manual human generation being time-consuming and costly, and artificial methods producing insufficient quality.
Innovation Solution
Simulating human behavior on electronic devices to generate network data traffic that mimics real-world traffic patterns, using a system comprising modules to control applications, devices, and network connections, and capturing and post-processing the data for machine-learning training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If human experts manually generate training data, then the quality of training data is high, but the process is time-consuming and costly
Solution Approach 1:
The patent uses automated bots to copy and simulate human behavior patterns for generating training data. Instead of manually creating each training sample, the system replicates human interaction patterns through automated agents that perform actions such as browsing, clicking, and navigating applications, thereby producing high-quality training data at scale without human time investment
Solution Approach 2:
The patent replaces the manual mechanical process of human experts generating training data with an automated electronic system. Bots equipped with machine learning models automatically generate training data by simulating user interactions with applications, substituting human manual operations with automated computational processes that maintain data quality while eliminating time constraints
2Productivity
If automated methods are used to generate training data, then time consumption is reduced, but the quality of training data becomes insufficient
Solution Approach 1:
The system copies authentic human behavior patterns by recording actual user interactions and using these recordings to train bots. The bots then replicate these observed behaviors to generate training data that maintains the same quality characteristics as manually created data, bridging the gap between automated efficiency and manual quality
Solution Approach 2:
The patent implements feedback mechanisms where generated training data is evaluated against quality metrics, and the bot generation process is adjusted based on this feedback. Machine learning models analyze the generated data and refine future data generation to maintain high quality standards while preserving automated efficiency
Data Source
Figure 1
Figure 2
AI summary
A method of generating network data traffic is described. The method comprises the following steps: - generating application use case data by means of an application use case module (18), wherein the application use case data is associated with at least one use case of at least one application running on at least one electronic device (14); - generating human behavior data by means of a human behavior module (20), wherein the human behavior data is configured to imitate human behavior when operating the at least one electronic device (14); - generating control data based on the application use case data and based on the human behavior data by means of an application execution module (22); and - generating network data traffic based on the control data by means of the at least one electronic device (14). Further, a training method for training a machine-learning module (12), a machine-learning module (12), and a system (10) for generating network data traffic are described.