Anomaly Detection Model Using Important Feature Extraction for IoT Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection technologies face challenges in low-specification IoT devices due to high computational complexity, particularly with autoencoder models and RaPP methods, which are not suitable for devices with limited processing power.
Innovation Solution
A low-complexity anomaly detection method that identifies important features using an autoencoder model and transmits this information to a low-specification device, allowing it to determine data anomalies with reduced computational power by employing models like isolation forest, PCA, SVM, DBSCAN, or normal distribution techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an autoencoder model is used for anomaly detection, then detection accuracy is improved, but computational complexity increases making it unsuitable for low-specification IoT devices
Solution Approach 1:
The patent segments the anomaly detection task into two parts: (1) an autoencoder model that identifies important features during a training phase, and (2) a simplified detection phase that uses only those important features with a low-complexity model. This segmentation allows the complex autoencoder to be used only when needed for feature identification, while the actual anomaly detection on IoT devices uses a simplified approach with reduced computational requirements.
Solution Approach 2:
The patent extracts only the important features identified by the autoencoder model from the complete dataset, discarding irrelevant features. This extraction process reduces the dimensionality of the data and allows the subsequent anomaly detection to be performed on a smaller, more manageable feature set, thereby reducing computational complexity while maintaining detection accuracy.
2Measurement precision
If a RaPP method is used for anomaly detection, then detection performance is improved, but computation amount increases
Solution Approach 1:
Instead of processing all features equally, the patent applies partial action by focusing computational resources only on the important features identified by the autoencoder. This selective processing reduces the overall computation amount required for anomaly detection while maintaining detection performance, as the critical features are analyzed in detail while non-critical features are either analyzed less thoroughly or not at all.
3Measurement precision
If all features are used for anomaly detection, then detection completeness is improved, but processing time increases
Solution Approach 1:
The patent extracts and isolates only the important features that contribute most to anomaly detection, separating them from the complete set of features. This extraction enables the system to achieve detection completeness by focusing on the most relevant features, while simultaneously reducing processing time by eliminating the need to analyze all features equally.
Solution Approach 2:
The patent changes the parameter of feature selection from using all available features to using only important features identified by the autoencoder. This parameter change in the feature set size and composition reduces the computational burden and processing time while maintaining the ability to detect anomalies effectively through the use of the most discriminative features.
Data Source
AI summary
Provided is an anomaly detection method performed by an electronic device. The method performed by an electronic device including one or more processors, a communication circuit which communicates with an external device, and one or more memories storing at least one instruction executed by the one or more processors may include: by the one or more processors, receiving target data for discriminating whether an anomaly occurs, in which the target data includes a value for each of a plurality of features; inputting a value for at least one important feature among the plurality of features into an anomaly detection model, in which the at least one important feature is determined by important feature information received from the external device; and determining whether the target data is abnormal based on an output of the anomaly detection model.