Memory-Less Anomaly Detection With Partial Model Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems require significant computing resources and data storage, which makes them vulnerable to malicious attacks and inefficient in terms of resource utilization.
Innovation Solution
Implementing a memory-less anomaly detection system that re-trains an inference model as needed, discards data after use, and performs partial re-training to minimize resource consumption and secure data from unauthorized access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data is stored for anomaly detection, then detection accuracy is improved, but data security deteriorates due to vulnerability to malicious attacks
Solution Approach 1:
The patent implements memory-less anomaly detection where training data is discarded after use rather than stored long-term. The system performs partial re-training as needed, using temporary data storage that is deleted after the re-training process completes. This approach maintains detection accuracy by allowing periodic re-training while eliminating security vulnerabilities associated with long-term data storage.
2Reliability
If full re-training of inference model is performed, then anomaly detection capability is improved, but computing resource expenditure increases
Solution Approach 1:
The patent implements partial re-training of the inference model, where only a portion of the model is re-trained rather than the entire model. This selective re-training approach updates the anomaly detection capability when needed while significantly reducing computing resource expenditure compared to full re-training of the complete model.
3Measurement precision
If large amounts of data are stored for training, then model accuracy is improved, but device complexity and security risk increase
Solution Approach 1:
The patent implements a memory-less architecture where training data is stored temporarily during the re-training process and then discarded. The system allows for periodic re-training using available data without requiring long-term storage infrastructure, thereby reducing device complexity while maintaining model accuracy through selective re-training operations.
Data Source
AI summary
Methods and systems for anomaly detection in a distributed environment are disclosed. To manage anomaly detection, a system may include an anomaly detector and one or more data collectors. The anomaly detector may detect anomalies in data obtained from one or more of the data collectors using an inference model. To perform anomaly detection, the inference model may require periodic re-training. Data collected from the one or more data collectors may be used to re-train the inference model as needed. Following anomaly detection and/or inference model re-training, the data may be discarded to remove the data from the anomaly detector.