paolocmo/satsec-decomposition

SatSec Grounded Objective-Decomposition Dataset

Version 2.0 is a leakage-controlled replacement for the original dataset used in A Controlled Candidate-Set Benchmark for Offline Satellite-Security Plan Decomposition (DOI 10.48550/arXiv.2607.26371). It contains 24 authored full decompositions and 83 mechanically derived next-step rows across 24 cases. There are 82 train rows and 25 test rows; the six test cases never occur in training.

Important v2 correction

The original version exposed prose describing the expected sequence inside the user prompt. Consequently, its decomposition and next-step results measured extraction and formatting rather than planning from a high-level objective. Do not use v1 scores as evidence for decomposition performance.

In v2 each input starts with only Objective: <one-line high-level objective>, followed by eight shuffled SPARTA candidates. Automated audits reject legacy sequence markers, future-step titles in next-step prompts, non-disjoint case splits, malformed grounding, and non-self-consistent gold plans.

The GNSS gold was also corrected: EX-0002 is PNT geofencing. Signal generation, overpowering, and controlled carry-off are represented by EX-0014.04 PNT Spoofing.

Schema

Column	Meaning
messages	system/user/assistant chat turns
type	decompose or next_step
case	source case identifier
root	corpus attack-tree root
step	1-based next-step index, otherwise null
split	case-disjoint train or test
dataset_version	2.0.0
prompt_policy	objective-only-v2

The raw source-of-truth JSONL is under raw/tuning_set.v2.jsonl. The superseded 110-row snapshot is retained only for historical audit under legacy/v1/tuning_set.jsonl; it must not be used for the reported v2 experiments.

Intended use and limitations

Use only for authorized, development-time testing on emulated or consented systems. Do not use it against operational or on-orbit assets. The dataset contains concrete identifiers and actions in assistant targets. Completion-only label masking does not guarantee that facts remain outside learned weights or that memorization cannot occur.

Exact-identifier scoring has one curated reference per case and may penalize valid alternative decompositions. SPARTA page checks establish identifier/name consistency, not the correctness of every semantic mapping; expert review remains necessary.

Loading

from datasets import load_dataset
ds = load_dataset("paolocmo/satsec-decomposition")

Licensing and disclosure

Dataset: CC BY 4.0. Grounding text is derived from public incident sources and SPARTA; no ECSS standard text is redistributed. Generative AI tools assisted with code scaffolding, dataset and manuscript review, and language revision. The authors verified the dataset construction, semantic mappings, experimental results, references, and final manuscript and take full responsibility for the work.

Authors

• João Paolo Cavalcante Martins Oliveira — Universidade Federal do Rio Grande do Norte (UFRN) and SETI Institute — ORCID 0000-0003-4117-953X
• Lucas Teske — TeskesLab — ORCID 0009-0002-8526-7662
• Paulo Matias — Universidade Federal de São Carlos — ORCID 0000-0002-6504-5141

Citation

If you use this dataset, please cite the paper and the dataset:

@misc{oliveira2026controlled,
  author        = {Oliveira, João Paolo Cavalcante Martins and Teske, Lucas and Matias, Paulo},
  title         = {A Controlled Candidate-Set Benchmark for Offline Satellite-Security Plan Decomposition},
  year          = {2026},
  eprint        = {2607.26371},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CR},
  doi           = {10.48550/arXiv.2607.26371},
  url           = {https://arxiv.org/abs/2607.26371}
}

@dataset{satsecdecomp2026,
  author    = {Oliveira, João Paolo Cavalcante Martins and Teske, Lucas and Matias, Paulo},
  title     = {SatSec Grounded Objective-Decomposition Dataset},
  year      = {2026},
  url       = {https://huggingface.co/datasets/paolocmo/satsec-decomposition},
  doi       = {10.57967/hf/9586},
  publisher = {Hugging Face}
}