Data Mining with Digital Fingerprinting - Challenges, Chances, and Novel Application Domains

Abstract

During the last decades, digital fingerprinting was used for hundreds of security-related applications. The main purpose relates to tracking and identification procedures for both users and tasks. The role of digital fingerprinting in data mining area became very important. As a key scale-out technology, thermal fingerprinting represents an experimental case study, which was introduced to show new application domains for fingerprinting-based profiling. We are now able to monitor all kind of sensor sources in a generic way. The concept is adoptable to hundreds of novel application domains in the IoT & smart metering context.

In this paper, we summarize key features of the thermal fingerprinting approach. The feasibility is demonstrated in a large scaled data centre testbed with typical sensor sources, e.g., temperature, CPU load behaviour, memory usage, I/O characteristics, and general system information. As a result, the approach generates two-dimensional unique and indexable patterns.

Besides this case study, we introduce several further use cases for this kind of sensor data fingerprinting. This includes data mining projects in the area of urban mobility profiling or innovative & lightweight weather forecast models, but also profiling capabilities in body area networks (health monitoring, fitness applications). Finally, we describe remaining challenges and critical security issues that still have to be solved.