HTTP Request Classification for Adaptive Cloud Service Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud service access management systems struggle with inefficiencies in managing access levels and application use, leading to data loss risks due to outdated blacklists and frequent updates, and the inability to allow beneficial service features while blocking inappropriate usage.
Innovation Solution
A machine learning-based system classifies HTTP requests and responses using a multiclass classifier to identify actions, enabling precise control over cloud service interactions, allowing or blocking requests based on action classifications, and adapting to changes in back-end APIs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If blacklists or rules databases are used to block cloud services, then data loss risk is reduced, but the system becomes outdated quickly and requires frequent costly updates
Solution Approach 1:
The system employs machine learning models that automatically analyze and classify cloud service APIs without requiring manual updates. The model self-adapts to API changes by learning from observed service behaviors, eliminating the need for frequent manual blacklist updates while maintaining reliable data loss prevention.
Solution Approach 2:
The patent replaces the manual mechanical process of updating blacklists and rules with an automated machine learning system. The ML model automatically detects and adapts to API changes in cloud services, substituting the inefficient manual update mechanism with an intelligent autonomous system that maintains reliability without requiring frequent human intervention.
2Object-affected harmful factors
If hard-coded rules are applied to block certain requests, then inappropriate usage is prevented, but beneficial service features cannot be allowed
Solution Approach 1:
The machine learning model applies different classification rules to different aspects of cloud service requests. It can identify and block harmful actions (such as unauthorized data exfiltration) while simultaneously allowing beneficial features (such as legitimate file sharing or collaboration functions) based on the specific context and characteristics of each request, rather than applying blanket blocking rules.
Solution Approach 2:
The system changes the parameters used for decision-making from simple blacklist matching to multi-dimensional ML classification. By analyzing multiple features and parameters of each request simultaneously, the system can distinguish between appropriate and inappropriate usage patterns, enabling selective blocking that preserves beneficial service features while preventing harmful actions.
3Measurement precision
If blacklists are frequently updated to reflect API changes, then blocking accuracy is maintained, but operational cost and time increase
Solution Approach 1:
The machine learning model operates continuously to monitor and classify cloud service requests, maintaining accurate blocking performance without interruption. Unlike periodic blacklist updates, the ML system provides continuous adaptation and classification, ensuring blocking accuracy is maintained at all times without requiring scheduled downtime for updates.
Solution Approach 2:
The patent replaces the time-consuming manual process of updating blacklists with an automated machine learning system that continuously adapts to API changes. The ML model automatically learns from new service patterns and updates its classification rules in real-time, eliminating the manual update cycle and maintaining high blocking accuracy without the associated time costs.
4Reliability
If cloud services are completely blocked, then data loss is prevented, but productivity and collaboration capabilities are reduced
Solution Approach 1:
The machine learning model applies granular classification to individual requests rather than blocking entire cloud services. It identifies and allows beneficial collaboration and productivity features (such as file sharing, document collaboration, and communication functions) while selectively blocking only the specific actions that pose data loss risks, thereby maintaining both security and productivity.
Solution Approach 2:
Instead of implementing complete blocking of cloud services, the system applies partial action by selectively classifying and controlling individual requests. The ML model allows sufficient service access to maintain productivity and collaboration while imposing restrictions only where necessary to prevent data loss, avoiding the excessive action of total blocking that would harm business operations.
Data Source
AI summary
A computer-implemented method for activity monitoring with respect to online services. The method can include accessing a machine learning multiclass classifier, the machine learning multiclass classifier representing HTTP network request features and associated actions with respect to interacting with websites, receiving an HTTP request, extracting a feature set from the HTTP request, determining a request action classification for the HTTP request, determining the request action classification comprising processing the feature set extracted from the HTTP request to the machine learning multiclass classifier to classify the HTTP request, and providing access to the HTTP request and request action classification via an application programming interface.