Neural Network Database Intrusion Detection via Change Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intrusion detection systems in databases face challenges in accurately monitoring and detecting unauthorized access, especially with evolving threats, as they struggle to efficiently process large amounts of network traffic and differentiate between normal and abnormal behavior.
Innovation Solution
The implementation of a neural network architecture, specifically a convolutional neural network, that tracks changes in a database's data table by generating matrices representative of these changes using background subtraction techniques, allowing for the identification of unauthorized access by distinguishing between normal and abnormal behavior patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intrusion detection systems monitor and analyze real-time network flow, then they can detect unauthorized intrusion, but they struggle to efficiently process large amounts of network traffic and provide accurate results
Solution Approach 1:
The patent replaces traditional mechanical intrusion detection systems with a neural network-based system. The neural network automatically learns patterns from training data and detects intrusions by analyzing matrix representations of database changes, eliminating the need for manual rule-based analysis and significantly improving both accuracy and processing efficiency.
Solution Approach 2:
The patent transforms database change data into matrix representations with specific dimensions (e.g., 28 dimensions representing different database operations). This parameter transformation allows the neural network to process complex database states in a standardized format, improving both detection accuracy and processing speed.
2Reliability
If neural network models with millions of parameters are used to extract hierarchies of features, then they can learn from large amounts of data, but the complexity of the system increases
Solution Approach 1:
The patent segments the intrusion detection task into distinct components: (1) generating matrix representations of database changes with specific dimensions, (2) training the neural network on these matrices, and (3) using the trained network for detection. This segmentation makes the complex system more manageable and allows for optimized processing at each stage.
3Adaptability or versatility
If convolutional neural network architecture is designed for any database, then it can obtain relationships between samples and labels, but the system requires significant computational resources for training and processing
Solution Approach 1:
The patent performs preliminary actions by pre-processing database changes into matrix representations before feeding them to the neural network. This pre-processing step organizes the data in a way that the convolutional neural network can efficiently process, reducing the computational burden during actual detection operations.
Data Source
AI summary
Aspects described herein may allow for the application of an artificial neural network architecture to identify intrusion in a database. Changes to components of a data table of the database may be tracked as a snapshot of the changes over a period of time. Any change in the data table may be associated with a user. Utilizing multiple snapshots, a background substitution technique may be utilized to generate a matrix of the changes to the data table over a period of time. A model having an artificial neural network architecture may utilize the matrix as an input set to identify the user as an unauthorized user accessing the database.