IoT Anomaly Detection Using PCA and Mahalanobis Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-spec IoT devices face challenges in implementing existing security systems due to high computation complexity, making it difficult to apply anomaly detection technologies like autoencoders effectively.
Innovation Solution
Anomaly detection method using principal component analysis (PCA) to search for the principal component axis maximizing variance, projecting data onto this axis, calculating Mahalanobis distance, and comparing it with a threshold to detect abnormal data, which reduces computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If autoencoder-based anomaly detection is applied, then detection accuracy is improved, but computation complexity increases making it inapplicable to low-spec IoT devices
Solution Approach 1:
The patent extracts only the essential components needed for anomaly detection from the complex autoencoder architecture. Instead of using deep neural networks with multiple hidden layers, it extracts the core functionality of learning normal data patterns and detecting deviations, implementing this through simplified statistical methods (mean, variance, Mahalanobis distance) that achieve the same detection goal with minimal computation.
Solution Approach 2:
The patent replaces expensive, complex computational models with cheap, simple statistical calculations. The autoencoder's complex weight matrices and non-linear transformations are replaced with basic arithmetic operations (mean calculation, variance computation, distance measurement) that can be executed efficiently on low-spec IoT devices with limited resources.
2Reliability
If existing security systems are implemented on low-spec IoT devices, then security reliability is improved, but device performance deteriorates due to high resource requirements
Solution Approach 1:
The patent changes the computational parameters from complex neural network operations to simple statistical parameters (mean, variance, standard deviation). This parameter transformation allows the system to maintain security reliability through accurate anomaly detection while dramatically reducing computational load, memory usage, and processing time, thus preserving device performance on low-spec IoT devices.
Data Source
AI summary
An anomaly detection method includes searching for one principal component axis by analyzing a normal data set collected in time series from a plurality of IoT devices by using a principal component analysis technique, setting a center point of the principal component, receiving a currently measured measurement data set from the plurality of IoT devices, acquiring a linear transformation data set having a plurality of projection points as elements by projecting a plurality of measurement data which is each element in the measurement data set onto the principal component axis, calculating a Mahalanobis distance between the projection point and the central point, and detecting whether or not data of the IoT devices is abnormal by comparing the Mahalanobis distance calculated for each element with a threshold.