Cloud Ransomware Detection via User Feedback Loops

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional ransomware detection systems in cloud storage systems often generate false positives, leading to unnecessary resource utilization and file loss, as they lack effective user confirmation and machine learning-based validation to accurately identify and restore ransomware-impacted files.

Innovation Solution

A cloud storage system that employs a ransomware analysis engine with machine learning capabilities, which uses user feedback to validate ransomware detection, performs tests on file features such as encryption status, naming patterns, and content analysis, and requests user confirmation to restore files to previous versions, thereby reducing false positives and improving detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional ransomware detection systems perform aggressive detection and prevention actions, then ransomware protection is improved, but false positives increase causing unnecessary resource utilization and file loss

Engineering Contradiction:
Improveransomware protectionVSAvoiddetection accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The system implements feedback loops where user confirmations and restoration outcomes are fed back into the machine learning model. The model continuously learns from confirmed ransomware cases and false positive corrections, improving detection precision while maintaining high reliability through iterative refinement of detection algorithms

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system performs preliminary detection and flags potential ransomware files before taking definitive prevention actions. By implementing a staged approach with initial detection, user confirmation, and then restoration, the system reduces false positives while maintaining protection effectiveness

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If user confirmation is implemented for ransomware detection, then false positives are reduced, but detection time and user burden increase

Engineering Contradiction:
Improvedetection accuracyVSAvoiddetection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system applies partial automation by requiring user confirmation only for borderline cases that fall within a confidence threshold range. High-confidence detections are automatically resolved without user intervention, while only uncertain cases require confirmation, thereby minimizing time loss while maintaining high detection accuracy

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system dynamically adjusts the confidence threshold parameter based on historical data and detection patterns. By changing the threshold parameter, the system can balance between requiring user confirmation and enabling automatic resolution, optimizing detection time while maintaining accuracy

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If machine learning-based validation is implemented, then detection precision is improved, but system complexity and computational resources increase

Engineering Contradiction:
Improvedetection precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system introduces an intermediary layer between traditional detection methods and final decisions. The machine learning model acts as a mediator that validates and refines detection results, providing probabilistic assessments that guide subsequent confirmation or restoration actions without requiring complete system redesign

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The detection system is segmented into multiple independent components: initial detection algorithms, machine learning validation layer, user confirmation interface, and restoration execution. This segmentation allows each component to be optimized independently and reduces overall system complexity by distributing functionality across modular units

Inventive Principle:
Principle #1Segmentation

4Reliability

If files are restored to previous versions upon detection, then data loss is prevented, but false restoration of legitimate files occurs

Engineering Contradiction:
Improvedata protectionVSAvoidfalse restoration
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The system uses feedback from user confirmations and restoration outcomes to continuously improve the detection model. By learning from false restoration cases, the system refines its detection criteria and reduces the likelihood of incorrect restorations while maintaining effective protection against actual ransomware

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3776312B1Service identification of ransomware impacted files
Publication Date: 2024.04.24 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3776312B1 patent drawingFigure 1
  • EP3776312B1 patent drawingFigure 2
  • EP3776312B1 patent drawingFigure 3

AI summary

A cloud storage server receives a request from a client device to store a file. The cloud storage server stores the file in a storage device of the cloud storage server. The cloud storage server determines features of the server-stored file and detects ransomware activity based on the features of the server-stored file. The cloud storage server sends a notification of the ransomware activity to the client device. The client device confirms the presence or an absence of the ransomware activity in the server-stored file. The cloud storage server updates an operation of the detection of the ransomware activity based on the received ransomware confirmation.