Bloom Filter Anomaly Detection for Millisecond Sensor Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods in industrial systems are inadequate for timely and reliable detection of anomalies, particularly in critical services, as they often require processing times exceeding the necessary milliseconds for effective action, and may unnecessarily shut down systems, impacting availability and safety.
Innovation Solution
A computer-implemented method using Bloom filters to compare real-time sensor data with predefined signatures indicative of anomalous events, allowing for fast detection and appropriate control actions, such as stopping or switching to backup systems, while minimizing downtime and ensuring accurate identification of critical versus non-critical anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional anomaly detection methods (genetic algorithms, principal component analysis) are used, then detection accuracy may be adequate, but processing time exceeds the required millisecond detection time
Solution Approach 1:
The patent creates simplified copies of anomaly signatures using Bloom filters. Instead of storing complete anomaly profiles, it creates compact bit array representations that can be quickly compared against sensor data, dramatically reducing processing time while maintaining detection capability
Solution Approach 2:
The patent segments anomaly detection into two distinct phases: (1) Bloom filter-based rapid screening to identify potential anomalies in milliseconds, and (2) detailed analysis phase for confirmed anomalies. This segmentation allows the system to meet real-time requirements for detection while performing comprehensive analysis only when necessary
2Reliability
If anomaly detection is performed continuously with high sensitivity, then detection reliability improves, but system shutdowns increase reducing availability
Solution Approach 1:
The patent applies different detection thresholds and analysis depths to different anomaly types and system contexts. Critical anomalies trigger immediate shutdowns, while non-critical anomalies allow continued operation with monitoring, optimizing both safety and availability based on local conditions
Solution Approach 2:
The Bloom filter performs a partial check first, identifying only the most likely anomalies with high confidence. Full anomaly verification and system shutdown decisions are performed only for cases where the Bloom filter indicates a genuine match, avoiding unnecessary shutdowns for false positives
Data Source
AI summary
Some embodiments relate to computer-implemented method for determining an anomalous event in a system. The invention includes defining a Bloom filter representing predefined signature strings each indicative of an anomalous event in the system. The invention includes receiving sensor data from, sensors of the system, indicative of operational parameters associated with the system. The invention includes determining, based on the sensor data, a current signature string indicative of current operation of the system, and comparing the current signature string to the predefined signature strings to determine whether there is an anomalous event in the system, where performing the comparison includes applying the Bloom filter to the current signature string.