An Assessment Methodology for Learning Mission Characteristics
European Conference on Cyber Warfare and Security · 15.06.2026 · 25 (1) · 528-536
DOI: 10.34190/eccws.25.1.4599 ↗Kurz gefasst
Fachartikel in der Zeitschrift European Conference on Cyber Warfare and Security, online erschienen am 15.06.2026. Verfasst von Joshua Meise und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
This paper presents a novel methodology for assessing the effectiveness of machine learning algorithms in deriving a protocol specification from a set of network traffic captures. Such specifications describe the data formats and mission characteristics associated with valid messages and form the basis for protecting military vehicles from malicious traffic. This paper enhances previous work, which automatically generates a hardware guard from a hand-written grammar, to produce a fully automated, end-to-end process that generates a parser, either in software or as a hardware guard, directly from mission training data sets. A hardware guard is realized as a Field Programmable Gate Array (FPGA) circuit block implementing a parser. The parser performs deep packet validation, checking that every message conforms to the associated grammar and rejecting invalid packets. Several modern machine learning algorithms and specification inference tools exist that can be leveraged to automatically infer a specification from a data capture. Unfortunately, to date, there has been a paucity of mission data sets to provide ground-truth, and the lack of a common set of metrics to assess learning algorithms. To this end, this paper introduces a parser repository that provides a ground-truth data set for a variety of formats and protocols, alongside equivalent protocol specifications expressed in several formalisms, namely BNF, Hammer, Daffodil, and Daedalus. Each set of equivalent specifications is provided with a common collection of true and false test vectors to validate their operation and benchmark learning performance using appropriate metrics. Specifications for standard formats, including JavaScript Object Notation (JSON), Universal Resource Locators (URLs), and Hypertext Transfer Protocol (HTTP), provide the basis for generic testing. The Micro Air Vehicle Link Protocol (MAVLink) and Space Packet Protocol (SPP) are used as examples of mission-oriented grammars. MAVLink is a byte-oriented protocol, while SPP is a bit-oriented protocol, leading to fundamental differences in the way that learning algorithms must operate to validate packets; the latter is also indicative of the complexities involved in parsing of SAE J1939 – used on ground vehicles – and MIL-STD-1553 – used on aircraft.
Zitieren
APA
Meise, J., & Taylor, S. (2026). An Assessment Methodology for Learning Mission Characteristics. European Conference on Cyber Warfare and Security, 25(1), 528-536. https://doi.org/10.34190/eccws.25.1.4599
Vancouver
Meise J, Taylor S. An Assessment Methodology for Learning Mission Characteristics. European Conference on Cyber Warfare and Security. 2026;25(1):528-536. doi:10.34190/eccws.25.1.4599
BibTeX
@article{Meise2026,
author = {Meise, Joshua and Taylor, Stephen},
title = {An Assessment Methodology for Learning Mission Characteristics},
journal = {European Conference on Cyber Warfare and Security},
year = {2026},
volume = {25},
number = {1},
pages = {528--536},
doi = {10.34190/eccws.25.1.4599}
}
EUVIMED