Client-Hosted Neural Networks for Adaptive Trust Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current approaches for determining user trustworthiness in network-based computing services focus on monitoring client behavior by a service provider, lacking a comprehensive and adaptive mechanism to detect abnormal behavior and adjust trust levels dynamically.

Innovation Solution

A system utilizing neural networks on both clients and an authority to establish and update baselines of acceptable behavior, allowing clients to infer normal or abnormal behavior and adjust reputation values based on observed actions, with an authority managing and integrating behavior data to recalibrate these baselines.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If service providers monitor client behavior to determine user trustworthiness, then security and trust detection capability is improved, but system complexity and resource consumption increase

Engineering Contradiction:

Improvetrust detection capabilityVSAvoidmonitoring system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system divides the monitoring function into two segments: a first neural network deployed on the client device that performs local behavior analysis, and a second neural network on the service provider's server that performs centralized analysis. This segmentation reduces the complexity burden on any single component while maintaining comprehensive trust detection capability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The first neural network on the client device performs preliminary behavior analysis locally before data is sent to the server. This preliminary action filters and pre-processes behavior data, reducing the amount of data that needs to be analyzed centrally and lowering the complexity of the server-side monitoring system.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If behavior baselines are updated dynamically to adapt to changing user patterns, then detection accuracy is improved, but system stability and reliability may deteriorate

Engineering Contradiction:

Improveabnormal behavior detection accuracyVSAvoidtrust baseline stability

Core Design Contradiction:

Measurement precisionVSStability of the object's composition

Solution Approach 1:

The system implements dynamic baselines through the second neural network that continuously learns and adapts to changing user behavior patterns over time. The baseline is not static but evolves dynamically, allowing the system to maintain high detection accuracy while adapting to legitimate behavior changes.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system employs feedback mechanisms where the second neural network on the server receives behavior data from the first neural network on the client, analyzes it, and updates the trust baseline accordingly. This feedback loop enables continuous improvement of detection accuracy while maintaining system stability through controlled adaptation.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12462153B2Behavior modeling using client-hosted neural networks

Publication Date: 2025.11.04 NVIDIA CORP

AI summary

Apparatuses, systems, and techniques to detect abnormal behavior on clients using one or more neural networks on said clients. In at least one embodiment, use of or behavior on one or more clients is analyzed by a first neural network to detect abnormal behavior compared to a baseline of accepted behavior, and said baseline of accepted behavior is revised over time by a second neural network based on behavior observed on said one or more clients.