Coreset-AI Technology is a data compression and anomaly detection platform developed by Shield-IoT, designed to address the challenge of analyzing large-scale IoT network traffic in real time. The technology is based on mathematical constructs called "coresets," which compress large datasets — from n data points down to log(n) — in real-time streaming mode, before applying existing AI/ML anomaly detection algorithms on the reduced dataset. This approach is described as problem-dependent data reduction (not generic compression), meaning it is specifically optimized for detecting network anomalies and removes noise in the process. Key technical characteristics: - Compresses data from 1 million to approximately 20 data points with near-zero loss of analytical fidelity - Supports fixed big datasets, distributed data, and streaming data - Reduces false positive rates compared to heuristic-based approaches (which can exceed 95% false positives) - Delivers anomaly detection results in minutes rather than hours or days - Applicable to IoT networks ranging from tens of thousands to millions of devices The underlying research spans over 15 years of academic work conducted at MIT and the University of Haifa, supported by over 70 academic papers and multiple patents. The technology has been benchmarked against RANSAC, a known anomaly detection algorithm, demonstrating lower error rates (false positives). Shield-IoT positions this as a core component of its IoT security management platform, targeting use cases where delayed detection of cyber-attacks or operational failures could result in disruption of physical services or financial harm.