Network Threat Response Using Reinforcement Learning Simulation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Network administrators face an overwhelming number of alerts and challenges in determining appropriate responses to increasing threats, especially with new attacks, and existing systems struggle to adapt to network changes and uncertainties in security analysts' responses.

Innovation Solution

A computer-implemented method using an intelligent agent trained with reinforcement learning to simulate threat propagation and determine effective responses by generating a model of the network, evaluating actions, and applying them to counter threats.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If signature-based detection is used to detect attacks, then detection accuracy for known attacks is improved, but the system cannot detect new attacks that have not been previously seen

Engineering Contradiction:
Improvedetection accuracyVSAvoidability to detect new attacks
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent combines signature-based detection and anomaly-based detection into a unified system. The anomaly detector identifies potential new threats by detecting deviations from normal behavior patterns, while the signature-based component provides accurate identification of known threats. This merging allows the system to achieve both high detection accuracy for known attacks and the ability to detect new, unknown attacks.

Inventive Principle:
Principle #5Merging (Combining)

2Adaptability or versatility

If anomaly-based detection is used to detect attacks, then the system can detect new attacks, but it requires extensive knowledge of normal network behavior which takes time to establish

Engineering Contradiction:
Improveability to detect new attacksVSAvoidtime to establish baseline
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs preliminary learning of normal network behavior patterns during an initial phase before full detection begins. This preliminary action establishes the baseline anomaly detection model in advance, so that when new attacks occur, the system can immediately detect them without requiring additional time to understand normal behavior. The learning phase captures typical network traffic patterns, which are then used for rapid anomaly detection.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If network administrators manually analyze and respond to each threat, then responses can be customized, but the overwhelming number of alerts makes this approach unsustainable

Engineering Contradiction:
Improveresponse qualityVSAvoidresponse speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system enables self-service automated response where the intrusion prevention system automatically analyzes detected threats, determines appropriate response actions, and executes them without requiring manual administrator intervention for each alert. The system serves itself by maintaining a playbook of response procedures and automatically selecting and executing the appropriate response based on the detected threat type and severity, thereby maintaining high response quality while achieving rapid automated responses to overwhelming numbers of alerts.

Inventive Principle:
Principle #25Self-service

4Productivity

If predetermined responses are stored for known threats, then response time is improved, but the system cannot adapt to new threats or changes in network configuration

Engineering Contradiction:
Improveresponse timeVSAvoidability to adapt to new threats
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The system maintains a dynamic playbook that can be automatically updated and adapted. While predetermined responses are stored for known threats to enable rapid response, the system continuously learns from new threats and network behavior changes, dynamically updating the playbook with new response procedures. This dynamic adaptation allows the system to maintain fast response times for known threats while simultaneously adapting to new threats and network configuration changes through continuous learning and automated playbook updates.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4324163B1Network protection
Publication Date: 2026.01.14 BRITISH TELECOM PLC
  • EP4324163B1 patent drawingFigure 1
  • EP4324163B1 patent drawingFigure 2
  • EP4324163B1 patent drawingFigure 3

AI summary

Mechanism for protecting a network comprising: receiving an indication that one or more anomalies determined to be associated with a threat have been detected within the network (310); generating a model of the network for simulating the propagation of the threat within the network (350); training an intelligent agent (360) to generate a response to the threat, the intelligent agent being trained using a reinforcement learning technique by using the model of the network to evaluate the effectiveness of taking different actions to counter the threat based on a simulated propagation of the threat within the model; using (370) the intelligent agent to determine a response to the threat, the response comprising one or more actions to be taken in relation to the network; and applying (380) the response to the network by causing the one or more actions to be taken in relation to the network.