Ensemble Machine Learning for Real-Time Application Execution Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional application execution control systems require frequent updates and consume significant resources, becoming inefficient as their databases grow, and struggle to manage dynamic changes in program behaviors effectively.
Innovation Solution
Implementing an ensemble of highly trained machine learning models that analyze a combination of static and dynamic features from programs and their environments in real-time, using a feature collector, discernment engine, and enforcement system to determine whether to allow program execution, leveraging models like neural networks, support vector machines, and logistic regression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static methods such as databases of signatures are used to determine program safety, then the system can control application execution, but the system requires frequent updates and consumes significant resources
Solution Approach 1:
The patent transitions from static signature databases to dynamic machine learning models that continuously learn and adapt to new program behaviors. The system collects runtime features and dynamically updates its classification models, eliminating the need for frequent manual database updates while maintaining high accuracy in execution control decisions.
Solution Approach 2:
The patent replaces the mechanical database update process with automated machine learning classification. Instead of manually maintaining and updating signature databases, the system uses trained models that automatically classify programs based on collected features, significantly reducing human intervention and resource consumption for updates.
2Measurement precision
If databases grow to improve detection capability, then the system can identify more program behaviors, but the management overhead and resource consumption increase significantly
Solution Approach 1:
The patent segments the detection system into multiple independent machine learning models, each trained on specific program behavior patterns. Instead of maintaining one large, complex database, the system divides detection into specialized models that can be independently trained, updated, and managed, reducing overall system complexity while improving detection precision.
Solution Approach 2:
The system implements self-service through automated feature collection and model training. The machine learning models automatically learn from collected program features without requiring manual database curation, reducing management overhead while continuously improving detection capabilities through self-directed learning.
3Reliability
If conventional application control systems are used, then execution control can be maintained, but the systems struggle to manage dynamic changes in program behaviors
Solution Approach 1:
The patent implements dynamic adaptation by using machine learning models that continuously learn from runtime program features. The system collects features during program execution and uses these dynamic inputs to adapt its classification decisions, enabling it to respond to new and evolving program behaviors while maintaining stable execution control through proven ML classification techniques.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Described are techniques to enable computers to efficiently determine if they should run a program based on an immediate (i.e., real-time, etc.) analysis of the program. Such an approach leverages highly trained ensemble machine learning algorithms to create a real-time discernment on a combination of static and dynamic features collected from the program, the computer's current environment, and external factors. Related apparatus, systems, techniques and articles are also described.