IoT Anomaly Detection Using Multi-Dimensional Device Grouping

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for monitoring large groups of IoT devices are resource-intensive, requiring numerous assessment modules to identify abnormal behavior, which becomes impractical with large deployments.

Innovation Solution

Assigning IoT devices to a multi-dimensional space with orthogonal dimensions, where each device is monitored by a subset of assessment modules, and combining outputs to identify non-compliant devices, using one-class classification and machine learning on frame headers to analyze time series data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If one uses a conventional monitoring system with one assessment module per IoT device, then each device can be monitored individually, but the resource consumption becomes extremely high when deploying large numbers of devices

Engineering Contradiction:
Improvedevice monitoring accuracyVSAvoidnumber of assessment modules
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the monitoring task by grouping IoT devices into clusters and assigning a single assessment module to each cluster rather than having one module per device. This segmentation approach maintains monitoring capability while significantly reducing the total number of assessment modules required, directly resolving the contradiction between monitoring accuracy and system complexity

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each assessment module is designed to serve multiple IoT devices within its cluster by analyzing aggregated behavioral patterns. The module performs universal monitoring functions across multiple devices, enabling one module to replace what would traditionally require many individual modules, thus reducing resource consumption while maintaining monitoring effectiveness

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If one increases the number of assessment modules to monitor more IoT devices, then coverage increases, but the system becomes less scalable and more resource-intensive

Engineering Contradiction:
Improvesystem scalabilityVSAvoidcomputational resource consumption
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent introduces a hierarchical dimension to the monitoring architecture by organizing devices into clusters at multiple levels. This dimensional organization allows the system to scale by adding clusters rather than adding individual assessment modules, improving scalability while reducing per-device resource consumption through aggregated analysis

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If one analyzes payload data from IoT devices to detect abnormal behavior, then detection accuracy improves, but privacy concerns arise and data security risks increase

Engineering Contradiction:
Improveabnormal behavior detection accuracyVSAvoidprivacy risk
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent extracts only the necessary behavioral metadata from device communications rather than analyzing the full payload data. By taking out only the essential features needed for anomaly detection (such as communication patterns, timing, and protocol behavior) while excluding sensitive payload content, the system maintains detection accuracy while eliminating privacy risks associated with payload analysis

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4078923B1Methods and apparatus to identify abnormal behavior within a set of internet-of-things devices
Publication Date: 2026.03.25 ORANGE SA
  • EP4078923B1 patent drawingFigure 1~2
  • EP4078923B1 patent drawingFigure 3
  • EP4078923B1 patent drawingFigure 4

AI summary

Methods and apparatus automatically identify which Internet-of-Things (loT) devices within a set are behaving in a manner non-compliant to a target behaviour. Each loT device (11) is assigned to a grid point in a notional m-dimensional space. A respective assessment module (2/3) is arranged to monitor behaviour of a group (12/13) of loT devices assigned to grid points that are aligned with one another at a respective position along the respective dimension, and to produce an output indicative of non-compliant behaviour in the event that the monitoring indicates that behaviour in the group of loT devices deviates from a behavioural model (15) of the loT devices of said group. An identification module (5) identifies at least one non-compliant loT device in the set of loT devices by combining outputs from the assessment modules assigned to the different dimensions of the space. The assessment modules may use trained machine-learning algorithms embodying a model of normal behaviour of the group of devices.