Adaptive Spatial Voting Grid for Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer security methods face challenges in detecting anomalies on memory or processor-restricted devices, particularly in identifying unknown-unknown behaviors without requiring extensive training data or labeled examples, leading to inefficiencies and inaccuracies in anomaly detection.
Innovation Solution
The implementation of adaptive Spatial Voting (SV) grids using Information Theoretic Methods for compressive sensing encoding, which enables self-initializing and self-aware anomaly detection by transforming data into numeric features for real-time analysis, eliminating the need for training data and traditional machine learning classifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning classifiers are used for anomaly detection, then detection accuracy may improve, but memory overhead and processing resource requirements increase significantly
Solution Approach 1:
The patent extracts only the essential features needed for anomaly detection (spatial voting grid positions and vote counts) rather than using full machine learning models. This extraction approach maintains detection capability while dramatically reducing memory requirements by storing only grid cell indices and vote counts instead of entire training datasets or model parameters.
Solution Approach 2:
The patent uses lightweight, disposable data structures (simple arrays storing grid cell indices and vote counts) instead of persistent, memory-intensive machine learning models. These minimal data structures can be easily created and discarded, providing anomaly detection functionality with negligible memory overhead compared to traditional classifiers.
2Reliability
If extensive training data is collected for anomaly detection, then detection reliability improves, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary action by pre-defining the spatial voting grid structure and encoding scheme before actual anomaly detection begins. This pre-processing creates a ready-to-use framework that eliminates the need for time-consuming training phases, allowing immediate deployment with reliable detection capability without extensive prior data processing.
Solution Approach 2:
The system performs self-service by automatically encoding incoming data into grid positions and updating vote counts without requiring external training or manual intervention. The anomaly detection mechanism is self-sufficient, using the predefined grid structure to immediately evaluate new data points and identify anomalies without additional processing time for training or model updates.
3Measurement precision
If complex anomaly detection algorithms are implemented, then detection precision improves, but device complexity and resource requirements increase
Solution Approach 1:
The patent segments the anomaly detection problem into discrete spatial voting grid cells, where each cell independently votes for anomaly presence. This segmentation transforms a complex continuous detection problem into simple discrete comparisons, maintaining precision while reducing algorithmic complexity to basic grid position calculations and vote count increments.
Solution Approach 2:
The patent changes the detection parameter from complex feature space comparisons to simple grid cell vote counts. By transforming the detection metric from multidimensional feature analysis to univariate vote counting in predefined grid cells, the system achieves comparable precision with significantly reduced computational complexity suitable for resource-constrained devices.
Data Source
AI summary
A method of identifying anomalous behavior can include determining a first feature value and a second feature value of a series of numbers, identifying whether the determined first and second feature values map to a cell that is a part of a group based on spatial voting (SV) grid data, the SV grid data indicating an extent of each cell in a grid of cells, a number of rows and columns of cells in a group of cells, and, for each group of cells, a point in the center cell of the group of cells and a unique group number, and in response to identifying that the determined first and second feature values do not map to a cell of the grid of cells that is a part of a group, tagging the first and second feature values as corresponding to an anomalous behavior.