Method for Improved IoT Prognostics and Improved Prognostic Cyber Security for Enterprise Computing Systems


Telemetry sampling rates from enterprise servers and for IoT asset prognostic monitoring are often constrained by hardware limitations in physical transducers and A/D digitizing firmware, and by hardcoded firmware in data acquisition instrumentation. The sampling restrictions impose challenges in terms of training advanced pattern recognition for prognostic applications such as Prognostic Cyber Security in enterprise and cloud data centers, and prognostic health management for end-customer IoT critical assets. No matter how slow the inherent sampling rate capabilities are for monitored assets, this paper introduces a novel empirical systematic and rigorous process to produce arbitrarily high telemetry sampling densities from assets for which such high telemetry sampling densities are physically and electronically impossible. This capability is achieved with no hardware or firmware modifications in any of the critical assets being monitored, and hence is backward compatible with legacy assets already in operation.

Proceedings of the 2017 International Conference on Artificial Intelligence