Trust Scoring via Sensor Data for Cheating Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current matchmaking systems in online multiplayer video games are inaccurate in predicting cheating behavior and rely on static rules, leading to poor player experiences due to mismatches between legitimate and cheating players.

Innovation Solution

A machine learning-based system that uses historical data from client machines, including sensor data from handheld devices, to generate trust scores for players, distinguishing between human and non-human generated game control data to isolate cheating players during matchmaking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If static rules-based systems are used to identify cheating players, then the system is simple to implement, but the accuracy in predicting cheating behavior is poor

Engineering Contradiction:
Improveaccuracy in predicting cheating behaviorVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces static rules-based systems with machine learning models that dynamically analyze sensor data patterns. The system uses trained machine learning models to process accelerometer, gyroscope, and microphone data, substituting mechanical rule-based detection with adaptive algorithmic analysis that improves cheating detection accuracy while managing system complexity through efficient model deployment.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameters analyzed by examining multiple sensor data dimensions (acceleration patterns, gyroscope readings, microphone audio) rather than relying on single static rules. By analyzing temporal patterns and correlations across multiple parameters, the system achieves higher prediction accuracy without proportionally increasing overall system complexity.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If sensor data from handheld devices is collected to improve cheating detection, then the accuracy of identifying cheating players improves, but the device complexity and data processing requirements increase

Engineering Contradiction:
Improveaccuracy of identifying cheating playersVSAvoiddata processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system extracts only the most relevant sensor data features needed for cheating detection (specific accelerometer patterns, gyroscope readings, and microphone audio characteristics) rather than processing all possible device data. This selective extraction maintains high detection accuracy while reducing data processing complexity and computational requirements.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Sensor data is pre-processed and features are extracted on the client device before transmission to the server. The system performs preliminary filtering, normalization, and feature extraction locally, which reduces the complexity of server-side processing while maintaining the accuracy needed for effective cheating detection.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If machine learning models are trained on historical sensor data, then the system adapts to changing player dynamics, but the computational resources and time required increase

Engineering Contradiction:
Improveadaptability to changing player dynamicsVSAvoidtraining time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

Machine learning models are trained in advance on historical sensor data from legitimate and cheating players before deployment. This preliminary training allows the system to adapt to various player dynamics and cheating patterns beforehand, enabling fast real-time inference without requiring extensive training time during actual gameplay matching.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses copied and replicated training data from historical gameplay sessions to train robust machine learning models. By utilizing abundant historical sensor data copies, the models achieve high adaptability to different player behaviors and cheating strategies without requiring continuous real-time training, thus minimizing time loss.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4025312B1Machine-learned trust scoring based on sensor data
Publication Date: 2025.01.01 VALVE CORPORATION
  • EP4025312B1 patent drawingFigure 1
  • EP4025312B1 patent drawingFigure 2
  • EP4025312B1 patent drawingFigure 3

AI summary

A trained machine learning model(s) is used to determine scores for user accounts registered with a video game service, and the scores are used to match players together in multiplayer video game settings. For example, sensor data received from client machines can be input to the trained machine learning model(s), and the model(s) generates scores as output, which relate to probabilities that the game control data received from those client machines was generated by handheld devices, as opposed to having been synthesized and/or modified using software. In this manner, subsets of logged-in user accounts executing a video game can be assigned to different matches (e.g., by isolating non-human players from human players) based at least in part on the scores determined for those logged-in user accounts, and the video game is executed in the assigned match for each logged-in user account.