Efficient Pattern-Based Analysis of Network Flows in IoT Systems

The exponential growth of IoT deployments introduces new challenges in real-time network security and behavioral monitoring. Existing Intrusion Detection Systems (IDS) and activity analysis tools often face significant limitations when scaling to high-volume, high-speed IoT data streams. This paper proposes a novel framework that leverages advanced data series indexing methods, building upon our previous work and ULISSE’s variable-length subsequence indexing, to enable scalable, real-time pattern matching for IoT network flow analysis. We detail algorithmic steps for in-memory indexing, multidimensional flow analysis, and adaptive similarity search to enhance detection accuracy and efficiency. An experimental roadmap for validation on realistic IoT datasets is presented.

In Digital Technologies for Future Cities Special Session, Proc. of International Balkan Conference on Communications and Networking (BalkanCom ‘25), Piraeus, Greece