Cloud Resource Criticality Scoring Using Class-Specific ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing and assigning criticality scores to cloud resources is challenging due to their widespread use across multiple organizations and teams, leading to difficulties in identifying critical resources and expanding the attack surface for potential vulnerabilities.
Innovation Solution
A method involving machine-learning-based classifiers and rule-based heuristics to derive resource criticality scores for cloud resources by converting information samples into vectorized data, using ML-based sentence transformers and n-gram representations, and applying cloud resource class-specific scaling and normalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If cloud resources are assigned default criticality scores using simple rule-based methodology, then the assignment process is fast and easy to implement, but the precision and accuracy of criticality identification deteriorates
Solution Approach 1:
The patent segments cloud resource criticality assessment into multiple dimensions including business impact, security risk, compliance requirements, and operational dependency. Each dimension is evaluated separately using specialized algorithms, then aggregated to produce a comprehensive criticality score. This segmentation allows both simple rule-based approaches for quick assessment and more sophisticated ML-based analysis for improved precision.
Solution Approach 2:
The patent implements dynamic criticality scoring where scores are continuously updated based on changing cloud resource states, usage patterns, and security threats. The system adapts weighting factors and evaluation criteria in real-time, transitioning from static default scores to dynamic, context-aware assessments that improve accuracy while maintaining computational efficiency.
2Measurement precision
If cloud resources are assigned detailed criticality scores using ML-based classifiers, then the precision of criticality identification is improved, but the device complexity and computational requirements increase
Solution Approach 1:
The patent applies preliminary filtering and feature selection before deploying ML-based classifiers. Common patterns and obvious critical resources are identified using simple rules first, allowing the system to focus computational resources on ambiguous or high-stakes cases. This preliminary action reduces the overall complexity while maintaining high precision for critical assessments.
Solution Approach 2:
The patent introduces intermediary processing layers including feature extraction modules, data normalization components, and hybrid scoring mechanisms that bridge simple rule-based methods and complex ML classifiers. These intermediaries simplify input data for ML models and interpret their outputs, reducing the complexity burden on the overall system while preserving accuracy benefits.
3Reliability
If cloud resources are assigned criticality scores based on comprehensive attributes, then the reliability of security functions is improved, but the loss of time for data processing increases
Solution Approach 1:
The patent implements periodic re-assessment of cloud resource criticality scores rather than continuous real-time analysis. Critical resources are evaluated more frequently, while less critical ones are assessed periodically. This periodic action maintains security reliability for important resources while reducing overall data processing time and resource consumption.
Solution Approach 2:
The patent applies different levels of analysis depth to different cloud resources based on their criticality level and attribute importance. High-criticality resources receive comprehensive multi-attribute analysis, while lower-criticality resources use streamlined assessment. This local quality approach ensures reliability where needed while minimizing processing time for less critical cases.
Data Source
AI summary
In an aspect, a machine-learning (ML)-based classifier or regressor associated with a respective cloud resource class by is trained inputting information samples and resource criticality scores for the respective cloud resource class as training data. In a further aspect, the ML-based classifier or regressor is further utilized to assign a resource criticality score to a particular cloud resource in the respective cloud resource class.