Strategic Technology Roadmapping in Maritime Sensing LLM-Based Zero- Shot Classification and NIST-CSF Mapping
Bildiriler > Tam Metin Bildiri · 2025
Yazarlar: Haydar YALÇIN
NATO SET-351 Research Specialists Meeting, İtalya, La Spezia, Uluslararası, Davetli Konuşmacı, Tam Metin Bildiri, 2025-2025, ss 1-11
Özet
<p>This proposal delineates a novel methodology for constructing a strategic technology roadmap pertinent to intelligent maritime sensing systems. It seamlessly integrates recent advancements in large language models (LLMs), zero-shot classification, and established cybersecurity standards. Specifically, we harness the analytical prowess of LLMs in conjunction with the National Institute of Standards and Technology Cybersecurity Framework (NIST-CSF) to orchestrate a perspicuous mapping of nascent technologies onto mission-critical functionalities within NATO’s multi-domain maritime sensing theatres. The inherently dynamic and frequently contested nature of maritime domains, from seabed to celestial sphere, engenders escalating imperatives for adaptive, resilient, and intelligent sensing technologies. As maritime unmanned systems (MUS) evolve through their amalgamation with AI, quantum, and full-spectrum sensing platforms, an overarching exigency arises for the systematic prognostication and assessment of technological trajectories. This proposal directly addresses this lacuna by deploying LLM-based analytical instruments to synthesise knowledge gleaned from patent repositories and scholarly research, thereby furnishing sagacious insights for NATO's strategic foresight and capability conceptualisation. Patent data will be meticulously culled from principal international repositories (WIPO, USPTO, EPO), whilst scholarly publications will be assiduously aggregated solely from Web of Science.This research endeavours to employ LLM-powered zero-shot classification to judiciously extract and categorise pertinent technologies from global patent and academic datasets. To align we will use LLM to identified technologies with the NIST-CSF's quintet of seminal cybersecurity functions: Identify, Protect, Detect, Respond, and Recover. We will use data analytics togenerate a visually intuitive and empirically driven technology roadmap, spanning the quinquennium 2025–2035, incorporating Technology Readiness Level (TRL) maturation estimations and trans-national innovation currents. We posit a domain-specific prompting paradigm that empowers LLMs (e.g., LLaMA 3 or GPT-4) to execute zero-shot classification of documents, allocating each entry to apposite NIST-CSF categories sans the prerequisite for supplemental supervised training. Through the judicious fusion of AI-enabled linguistic models and robust cybersecurity frameworks, this proposal introduces a methodology characterized by its intellectual rigour and strategic applicability, designed to augment NATO’s enduring capabilities within the maritime domain. The proposed framework possesses considerable potential to evolve into a shared instrumentality for technology reconnaissance, innovation appraisal, and multilateral research prioritisation.</p>