IoT Communication Anomaly Detection Model Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for detecting abnormal communication in IoT devices often misclassify normal communication as abnormal due to insufficient learning data, leading to increased operational loads and uncertainty about when to relearn models.
Innovation Solution
A detecting device that collects communication information, learns characteristics of this data, and relearns models when the number of detected abnormalities exceeds a threshold, improving accuracy and reducing operator load by dynamically adjusting learning parameters and thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If a model learns communication characteristics using a small amount of communication information, then the learning process is fast and resource-efficient, but the detection accuracy deteriorates because communication patterns cannot be fully covered
Solution Approach 1:
The system dynamically adjusts the learning process by implementing automatic relearning triggered when detection accuracy falls below a threshold. The model transitions from a static state to a dynamic state where it continuously adapts by relearning communication patterns, resolving the contradiction between fast initial learning and sustained high accuracy.
Solution Approach 2:
The system incorporates feedback mechanisms where detection results are monitored and fed back to trigger relearning when necessary. The abnormal communication detection accuracy serves as feedback to determine whether the model needs to relearn, creating a closed-loop system that maintains accuracy without requiring continuous manual intervention.
2Measurement precision
If the model continuously relearns to improve detection accuracy, then the detection accuracy improves, but the operation load and system complexity increase
Solution Approach 1:
The system performs self-service by automatically determining when relearning is needed based on detection accuracy thresholds. The model monitors its own performance and triggers relearning autonomously without requiring external intervention, thereby improving accuracy while minimizing unnecessary relearning operations and system complexity.
Solution Approach 2:
The system changes operational parameters dynamically by adjusting the relearning trigger based on detection accuracy thresholds. Instead of continuous or fixed-interval relearning, the system modifies its behavior based on performance parameters, relearning only when accuracy falls below the threshold, thus balancing accuracy improvement with operational efficiency.
3Ease of operation
If the system operates without relearning to maintain simplicity, then the operation load remains low, but the detection accuracy deteriorates when encountering unknown communication patterns
Solution Approach 1:
The system takes preliminary action by establishing detection accuracy thresholds and automatic relearning triggers in advance. This preliminary configuration allows the system to operate simply during normal conditions while being prepared to automatically improve reliability when needed, without requiring complex real-time decisions or manual intervention.
4Measurement precision
If manual determination of relearning timing is implemented, then relearning can be precisely controlled, but the operation load on system operators increases
Solution Approach 1:
The system eliminates manual determination of relearning timing by implementing self-service automation. The model automatically monitors detection accuracy and triggers relearning when thresholds are breached, achieving precise relearning timing control without requiring operator intervention or increasing workload.
Data Source
AI summary
A detecting device includes a memory, and processing circuitry coupled to the memory and configured to collect communication information from a communication device, have a model learn a characteristic of the communication information by the communication device using the communication information collected for each of the communication devices, and input communication information on a detection target to the model, detect whether the communication information on the detection target indicates abnormal communication on the basis of an output result from the model, and have the model relearn at the learning when the number of detected abnormalities about the communication information during a predetermined evaluation period exceeds a first threshold value.