Ensemble Model for Malicious Communication Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods face challenges in accurately specifying malignant communication destinations as attackers frequently change these, leading to inadequate model accuracy due to insufficient training data.

Innovation Solution

A calculation device and method that generates a mixed model by weighting and combining multiple training models, using ensemble learning to improve the accuracy of malignant degree calculations for communication destinations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a model is generated using only the latest malignant communication destination data, then the model can be updated quickly to reflect current threats, but the model accuracy is insufficient due to lack of sufficient training information

Engineering Contradiction:
Improvemodel accuracyVSAvoidtime to update model
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system maintains a time-series database of historical malignant communication destination data, performing preliminary data collection and storage across multiple time points. This allows the model to access sufficient training information from past data without requiring real-time updates, thereby maintaining accuracy while avoiding the time loss of continuous model retraining.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system combines multiple data sets from different time periods into a comprehensive training data set. By merging historical data with current data, the model receives sufficient training information across various time points, improving accuracy without relying solely on the latest data which may be insufficient on its own.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If multiple models are trained on different data sets, then the model accuracy improves through ensemble learning, but the system complexity increases

Engineering Contradiction:
Improvemalignant degree calculation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system divides the training data into multiple time-based data sets (first time period, second time period, etc.) and generates separate models for each. This segmentation allows each model to specialize in specific time period characteristics, improving overall measurement precision while maintaining manageable individual model complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system creates a universal evaluation mechanism that can handle multiple models generated from different data sets. The malignant degree calculation unit universally evaluates communication destinations by synthesizing results from multiple models, achieving high precision without requiring separate evaluation systems for each model.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If attackers frequently change malignant communication destinations, then the threat landscape becomes more dynamic and harder to track, but conventional single-model approaches cannot adapt quickly enough

Engineering Contradiction:
Improveadaptability to changing threatsVSAvoiddetection accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The system implements a dynamic multi-model architecture where models are trained on data from different time periods and can be selectively applied based on the characteristics of the communication destination being evaluated. This dynamic approach allows the system to adapt to changing threat patterns while maintaining detection accuracy through the collective intelligence of multiple specialized models.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3839781B1Calculation device, calculation method, and calculation program
Publication Date: 2024.09.11 NIPPON TELEGRAPH & TELEPHONE CORP
  • EP3839781B1 patent drawingFigure 1
  • EP3839781B1 patent drawingFigure 2
  • EP3839781B1 patent drawingFigure 3~4

AI summary

A calculation device (10) receives input of a plurality of pieces of training data including a communication destination known to be malignant as data. The calculation device (10) generates a model that calculates a malignant degree of an input communication destination from each piece of the training data. The calculation device (10) gives weight to each of the models, and generates a mixed model using the model and the weight. The calculation device (10) calculates a malignant degree of a communication destination unknown whether the communication destination is malignant using the mixed model.