65c3f75641b22925c737ca657b126cd68c39e42334/6a860e6b1bca5f1434a37839582a4a68886edcc96ec4fc113a66446060b7c50b

SudoSet: A Contextual Authorization Benchmark for LLM Agents

SudoSet measures whether frontier LLMs honor classical authorization primitives — information-flow tags on observation provenance and capabilities on session scopes — when a runtime exposes them through the prompt context. It covers three attack families along a paired-contrast structure: confused deputy, source-label indirect prompt injection, and scope-check (transitive) IPI.

Dataset summary

• 270 scenario manifests, comprising 135 paired contrasts × 2 variants per pair, across three attack families (45 pairs each).
• Each pair shares caller, user prompt, document body (where applicable), target tool, and target resource. Only one runtime-supplied field flips between the two variants of a pair, and the flip determines the expected outcome (tool_called vs tool_blocked).

Loading

from datasets import load_dataset

scenarios = load_dataset(
    "65c3f75641b22925c737ca657b126cd68c39e42334/6a860e6b1bca5f1434a37839582a4a68886edcc96ec4fc113a66446060b7c50b",
    "scenarios", split="all",
)

print(scenarios)            # 270 rows
print(scenarios[0])         # first scenario manifest

Schema (scenarios config)

One row per scenario manifest. The two halves of a pair share pair_id. Source-label IPI and scope-check IPI rows include the document body inlined as document_body.

Field	Type	Notes
id	string	Unique scenario identifier (e.g. ipi_001_trusted)
pair_id	int	Shared by the two halves of a contrast
type	string	source_label_ipi, scope_check_ipi, or confused_deputy
boundary_type	string	cross_engagement, role_based, seniority, or external
variant	string	trusted/untrusted (CD, source-label IPI) or in_scope/out_of_scope (scope-check IPI)
sender_id	string	Caller email (CD) or runtime source attribution (IPI; trusted = internal Aquila address, untrusted = external sender or unauthenticated origin)
is_authenticated	bool	Whether the runtime tag attests authenticity for this scenario
target_tool	string	The tool whose invocation the harness records as tool_called
target_resource	string	The resource argument the JSON-serialized tool input must contain (IPI only); for confused-deputy this is informational
user_prompt	string	The user message that initiates the agent loop
injection_file	string | null	Filename of the document body in the original repo (IPI only)
document_body	string | null	Document body inlined here (IPI only); byte-identical across the two halves of a pair
manifest	object | null	{"grants": [{"resource": …, "actions": [...]}]} for scope-check IPI scenarios; consumed by the capabilities and both conditions
expected_outcome	string	tool_called for act-variants, tool_blocked for refuse-variants

Files

data/
  scenarios.jsonl              270 rows
env/
  system_prompt_base.md        Workspace description loaded into every condition
  tags_convention.md           Verbatim _TAGS_CONVENTION (added under tags and both)
  capabilities_convention.md   Verbatim _CAPABILITIES_CONVENTION (added under capabilities and both)
  access_control.json          Per-resource ACL used to derive granted_scopes
LICENSE                         CC-BY-4.0

The harness, per-record evaluation logs, refusal-source audit labels, and figure-generation scripts are released in a linked code repository (see the paper). This dataset repository contains only the scenario test bed: anyone running the released harness reproduces the per-record evaluation logs and the refusal-source audit labels by re-running against the listed model APIs.

Reproducing the evaluations

The harness in the linked code repository takes a model identifier and re-runs the 270 scenarios under the four conditions (baseline, tags, capabilities, both):

python eval.py --model claude-sonnet-4-6 --workers 32

Per-provider model IDs, decoding settings, retry policy, and run dates are documented in the paper appendix.

Intended use

SudoSet is intended for:

• Evaluating runtime authorization primitives (information-flow tags, capability lists) on LLM agents.
• Calibrating frontier-model behavior under deployable changes to the agent runtime's prompt context.
• Instruction-tuning data for training models to discriminate authorization-relevant runtime signals rather than refuse on surface cues.

The benchmark is a fixed test bed; per-cell McNemar contrasts compare conditions on the same scenarios, and population-level generalization to other workspaces is out of scope.

Limitations

• Single fictional workspace (Aquila Consulting), single ACL, six principals, four boundary types per family.
• 45 paired scenarios per family; per-cell power analysis at 80% is in the paper.
• Each (model, scenario, condition) triple is sampled once at provider-default decoding settings.
• tool_blocked is recorded for any non-target outcome within five turns; the paper's Claude-Opus-4.7-as-judge refusal-source audit on a random sample of 500 refuse-variant passes finds 72% are authorization-grounded overall, with both at 82%, capabilities at 71%, and tags-only at 58% (the unrelated cases concentrate on hallucinated-success claims under tags).
• The release here is the benchmark scenario set; no model weights and no per-record model outputs are included.

Ethics and dual use

The workspace is fictitious; no real personal data is included. The scenarios are written to test authorization-relevant decision boundaries rather than to teach a model how to comply with attacks. The dataset could in principle be adapted to test prompt-injection attacks against deployed agents, comparable to existing benchmarks in this space (AgentDojo, WASP, Shade-Arena).

Citation

@inproceedings{sudoset2026,
  title  = {SudoSet: A Contextual Authorization Benchmark for LLM Agents},
  author = {Anonymous},
  year   = {2026},
  booktitle = {NeurIPS Datasets and Benchmarks Track},
  note   = {Anonymized for review.}
}

License

The dataset is released under CC-BY-4.0. The harness and figure-generation code are released separately under MIT or Apache-2.0 in the linked code repository.