Add ASTRIDE diagrams, annotations, and generation trial outcomes

CITATION.bib

@@ -0,0 +1,7 @@
+@misc{astride_gen_dataset,
+  title = {ASTRIDE: Diagram-Grounded Threat-Modeling Dataset},
+  year = {2026},
+  howpublished = {Hugging Face Datasets},
+  url = {https://huggingface.co/datasets/lambdaeranga/astride},
+  note = {Version v0.1.0}
+}

README.md

@@ -0,0 +1,143 @@
+---
+pretty_name: ASTRIDE — Diagram-Grounded Threat-Modeling Dataset
+language:
+- en
+size_categories:
+- n<1K
+modalities:
+- image
+- text
+tags:
+- synthetic
+- threat-modeling
+- agentic-ai
+- architecture-diagrams
+- vision-language-models
+- evaluation
+configs:
+- config_name: comparison
+  default: true
+  data_files:
+  - split: evaluation
+    path: data/comparison/*.parquet
+- config_name: development
+  data_files:
+  - split: examples
+    path: data/development/*.parquet
+- config_name: trials
+  data_files:
+  - split: evaluation
+    path: data/trials/*.parquet
+---
+
+# ASTRIDE: Diagram-Grounded Threat-Modeling Dataset
+
+Version **v0.1.0** contains synthetic architecture diagrams, operating facts, candidate threat annotations, mitigation recommendations, and generation trial outcomes. It supports research on constructing datasets for ASTRIDE-style vision–language threat analysis.
+
+The dataset accompanies the draft **An Agentic Framework for Constructing and Evaluating Multimodal Threat-Modeling Datasets**. It is distinct from the corpus used in the original ASTRIDE threat-analysis paper. A publication reference will be added when available.
+
+## Collections
+
+| Configuration | Split | Records | Candidate threats |
+|---|---|---:|---:|
+| `comparison` (default) | `evaluation` | 48 diagrams | 154 |
+| `development` | `examples` | 3 diagrams | 14 |
+| `trials` | `evaluation` | 54 generation attempts | Not applicable |
+
+There are **51 diagram records and 168 candidate threats** in total. The controlled experiment uses only the 48 comparison diagrams and 154 threats. The three earlier development examples used a different prompt and recovery history and must not be pooled into that experiment. The 36-record authored template pilot is not included. No training split or verified threat-detection benchmark is provided.
+
+## Load diagrams and annotations
+
+```python
+from datasets import load_dataset
+
+repo = "lambdaeranga/astride"
+ds = load_dataset(repo, "comparison", split="evaluation", revision="v0.1.0")
+example = ds[0]
+image = example["image"]  # PIL image; PNG bytes are embedded in Parquet
+facts = {k: example[k] for k in ("title", "purpose", "assumptions", "controls")}
+architecture = example["architecture"]
+threats = example["threats"]
+mitigations = example["mitigations"]
+
+examples = load_dataset(repo, "development", split="examples", revision="v0.1.0")
+trials = load_dataset(repo, "trials", split="evaluation", revision="v0.1.0")
+```
+
+Images load without any files from the author's workstation. Every diagram is also provided as a PNG, SVG, and Mermaid source under `images/`. Portable JSONL records under `records/` use image paths relative to this repository.
+
+For VLM input, use the **image plus operating facts**. Canonical architecture labels, threats, mitigations, and generation-condition identifiers are targets or study metadata; do not leak them into the prediction input.
+
+## Record fields
+
+- `id`, `collection`, `trial_id`: stable record and trial identifiers.
+- `domain`, `control_profile`, `generation_condition`: experiment grouping metadata.
+- `topology_family_id`: typed graph-isomorphism group across both diagram collections, not a semantic-family reference label.
+- `image`, `image_path`: embedded image and repository-relative PNG path.
+- `title`, `purpose`, `assumptions`, `controls`: explicit operating facts.
+- `architecture`: trust zones, components with stable IDs/kinds, and directed data flows.
+- `threats`: category, attacker/preconditions, referenced components/flows/boundaries, evidence, impact, and exposed/reduced/unknown assessment.
+- `mitigations`: enforcement components, recommendation, proposed verification, existing-control effect, and residual risk.
+- `scope_notes`: generator-supplied limitations; these notes can themselves contain errors.
+- `annotation_status`: `candidate` for all records; no automatic admission to training is implied.
+
+`schema.json` records the exact Hugging Face feature schemas. Threat IDs join to mitigation `threat_id` values within each record. Trial rows use `record_id` to reference valid comparison examples; this field is empty for six unsuccessful attempts.
+
+## Construction
+
+The comparison assigned 18 requests to each of three conditions: single-prompt generation, decomposition without semantic review, and the full workflow. Requests crossed three domains (customer support, finance operations, incident response), three control profiles (weak, guarded, unspecified), and two repetitions. Generation used `gpt-5.4-mini-2026-03-17`, common structural validators, graph constraints, and annotation policy. Shared ceilings were three rounds, 12 logical calls per trial, and 6,000 output tokens per call; realized compute differed.
+
+| Condition | Valid / assigned | Logical calls | Reported input + output tokens |
+|---|---:|---:|---:|
+| Single prompt | 13/18 | 41 | 158,891 |
+| Without semantic review | 17/18 | 80 | 194,848 |
+| Full workflow | 18/18 | 117 | 320,626 |
+
+Failures and duplicate topologies were retained. One full-workflow trial exhausted repair while retaining a structurally valid candidate, which remains included. Validity means passing implemented structural checks, not semantic correctness. Trial data include failed attempts in cost/yield denominators; logical calls exclude SDK transport retries. No actual billing or evaluation-session cost is provided.
+
+The study used AI-assisted annotation evaluation, but detailed ratings are **not part of this dataset-only release**. No security ground truth, omissions recall, executed mitigation effectiveness, or downstream learning gain is established by the released data.
+
+## Repository layout
+
+```text
+README.md
+CITATION.bib
+release.json
+release_manifest.json
+schema.json
+data/
+  comparison/evaluation-00000-of-00001.parquet
+  development/examples-00000-of-00001.parquet
+  trials/evaluation-00000-of-00001.parquet
+records/
+  comparison.jsonl
+  development.jsonl
+  trials.jsonl
+images/
+  comparison/*.{png,svg,mmd}
+  development/*.{png,svg,mmd}
+```
+
+The release intentionally excludes implementation source, resolved generation prompts, raw model-call traces, intermediate drafts, and detailed evaluation ratings/packets. Protocol and rating hashes in `release.json` are lineage identifiers; the corresponding private artifacts are not accessible through this release. Full experimental reproduction is therefore not supported by this package alone.
+
+## Intended use and limitations
+
+Use these candidates to inspect synthetic diagram/annotation structure, develop annotation checks, and design prospective VLM training experiments. Known defects include unsupported threats, overstated certainty, partial mitigations, and questionable graph semantics. These annotations must not be treated as verified vulnerabilities or automatically correct training targets. Suggested verification procedures were not executed. The diagrams do not represent actual customers, payments, or incident records.
+
+Some diagrams share typed topology. There are 44 typed-isomorphism groups across the 51 diagrams, and semantic duplication can remain between different groups. Keep related graph families and all derived renderings together in any future split. The current `evaluation` split identifies a generation-study collection, not an independently validated benchmark.
+
+## License and citation
+
+**License: unspecified.** No license has been selected for this release. Check the applicable version's dataset card before reuse.
+
+Use `CITATION.bib` and identify the exact release tag or Hub revision used:
+
+```bibtex
+@misc{astride_gen_dataset,
+  title = {ASTRIDE: Diagram-Grounded Threat-Modeling Dataset},
+  year = {2026},
+  howpublished = {Hugging Face Datasets},
+  url = {https://huggingface.co/datasets/lambdaeranga/astride},
+  note = {Version v0.1.0}
+}
+```

data/comparison/evaluation-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca163b1aa6377f6733b46a69b0bd47acb9eb55938192ceb70a10f21966bf4151
+size 2982191

data/development/examples-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:254af98e07912e2646f0a7797a56714f217de6607454a5a00016a90aeee9fb00
+size 183663

data/trials/evaluation-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f7400ab3be5fcda2a2da4b14c96652d82c741d4d50f44b77fccbc761aa3a148
+size 8765

images/comparison/E001.mmd

@@ -0,0 +1,17 @@
+flowchart TB
+    subgraph Z_Z1["External source"]
+        C1["C1: Invoice documents"]
+    end
+    subgraph Z_Z2["Invoice processing"]
+        C2("C2: Invoice intake agent")
+        C3[("C3: Payment draft record")]
+    end
+    subgraph Z_Z3["Approval and execution"]
+        C4("C4: Approval reviewer")
+        C5("C5: Payment execution service")
+    end
+    C1 -->|"F1: invoice documents and supporting evidence [B1]"| C2
+    C2 -->|"F2: structured invoice extraction and payment draft"| C3
+    C3 -->|"F3: payment draft for review [B2]"| C4
+    C4 -->|"F4: approved payment instruction"| C5
+    C5 -->|"F5: execution status and payment reference [B2]"| C3

images/comparison/E002.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["External Sources"]
+        C5["C5: External Help Documents"]
+        C6["C6: Support Ticket Intake"]
+    end
+    subgraph Z_Z2["Support Application"]
+        C1("C1: Support Agent")
+        C2("C2: Help Document Fetcher")
+        C3("C3: Response Drafting Tool")
+    end
+    subgraph Z_Z3["Account Data"]
+        C4[("C4: Customer Account Store")]
+    end
+    C6 -->|"F1: customer question and ticket text [B1]"| C3
+    C5 -->|"F2: help article content [B1]"| C2
+    C2 -->|"F3: retrieved document snippets"| C3
+    C1 -->|"F4: account lookup request [B2]"| C4
+    C4 -->|"F5: account details [B2]"| C1
+    C3 -->|"F6: draft response and citations"| C1

images/comparison/E003.mmd

@@ -0,0 +1,17 @@
+flowchart TB
+    subgraph Z_Z1["External Evidence"]
+        C1["C1: Evidence Intake"]
+    end
+    subgraph Z_Z2["Analysis Workspace"]
+        C2("C2: Analyst Agent A")
+        C3("C3: Analyst Agent B")
+    end
+    subgraph Z_Z3["Coordination and Case Data"]
+        C4("C4: Coordinator")
+        C5[("C5: Case Store")]
+    end
+    C1 -->|"F1: incident documents, logs, attachments [B1]"| C2
+    C1 -->|"F2: incident documents, logs, attachments [B1]"| C3
+    C2 -->|"F3: normalized findings, provenance notes [B2]"| C4
+    C3 -->|"F4: normalized findings, provenance notes [B2]"| C4
+    C4 -->|"F5: consolidated case records, response status"| C5

images/comparison/E004.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["Untrusted document intake"]
+        C1["C1: Supplier portal"]
+        C2("C2: Document intake service")
+    end
+    subgraph Z_Z2["Invoice processing"]
+        C3("C3: Invoice agent")
+        C4[("C4: Provenance store")]
+    end
+    subgraph Z_Z3["Approval and payment execution"]
+        C5("C5: Approval console")
+        C6("C6: Payment service")
+    end
+    C1 -->|"F1: invoice documents and supporting evidence"| C2
+    C2 -->|"F2: normalized documents plus provenance metadata [B1]"| C4
+    C4 -->|"F3: invoice record and provenance context"| C3
+    C3 -->|"F4: payment draft with extracted payee, amount, and justification [B2]"| C5
+    C5 -->|"F5: approved payment request"| C6
+    C6 -->|"F6: payment confirmation or remittance notice [B3]"| C1

images/comparison/E005.mmd

@@ -0,0 +1,15 @@
+flowchart TB
+    subgraph Z_Z1["External evidence sources"]
+        C1["C1: Evidence intake"]
+    end
+    subgraph Z_Z2["Trusted analysis and case handling"]
+        C2("C2: Analyst agent A")
+        C3("C3: Analyst agent B")
+        C4("C4: Coordinator")
+        C5[("C5: Case store")]
+    end
+    C1 -->|"F1: incident documents, logs, attachments [B1]"| C2
+    C1 -->|"F2: incident documents, logs, attachments [B1]"| C3
+    C2 -->|"F3: analyst findings, extracted indicators"| C4
+    C3 -->|"F4: analyst findings, extracted indicators"| C4
+    C4 -->|"F5: consolidated case record"| C5

images/comparison/E006.mmd

@@ -0,0 +1,20 @@
+flowchart TB
+    subgraph Z_Z1["External Inputs"]
+        C1["C1: Customer Ticket Intake"]
+        C2["C2: Help Document Source"]
+    end
+    subgraph Z_Z2["Support Application"]
+        C3("C3: Support Orchestrator")
+        C4("C4: Response Drafting")
+        C5("C5: Account Lookup Gateway")
+    end
+    subgraph Z_Z3["Customer Data Store"]
+        C6[("C6: Customer Account Store")]
+    end
+    C1 -->|"F1: customer issue text and attachments [B1]"| C3
+    C2 -->|"F2: external help articles and examples [B1]"| C3
+    C3 -->|"F3: lookup request for customer context"| C5
+    C5 -->|"F4: account query [B2]"| C6
+    C6 -->|"F5: account status, history, and profile fields [B2]"| C5
+    C3 -->|"F6: selected ticket context, retrieved help content, and account facts"| C4
+    C4 -->|"F7: draft support response [B1]"| C1

images/comparison/E007.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["External Intake"]
+        C1["C1: Supplier Upload Portal"]
+    end
+    subgraph Z_Z2["Trusted Application"]
+        C2("C2: Invoice Agent")
+        C3[("C3: Document Store")]
+        C4("C4: Payment Draft Service")
+    end
+    subgraph Z_Z3["Approval and Execution"]
+        C5[("C5: Approval Queue")]
+        C6("C6: Payment Executor")
+    end
+    C1 -->|"F1: Invoice document and supporting evidence [B1]"| C2
+    C2 -->|"F2: Parsed invoice content and extracted metadata"| C3
+    C2 -->|"F3: Billing details and payment draft fields"| C4
+    C4 -->|"F4: Draft payment request for review [B2]"| C5
+    C5 -->|"F5: Approved payment request"| C6
+    C6 -->|"F6: Execution status and reconciliation result"| C5

images/comparison/E008.mmd

@@ -0,0 +1,22 @@
+flowchart TB
+    subgraph Z_Z1["Untrusted intake"]
+        C1["C1: Ticket front end"]
+        C2("C2: Evidence uploader")
+    end
+    subgraph Z_Z2["Guarded agent workspace"]
+        C3("C3: Document filter")
+        C4("C4: Support assistant")
+        C5("C5: Retrieval gate")
+        C7("C7: Draft response service")
+    end
+    subgraph Z_Z3["Protected account store"]
+        C6[("C6: Customer account store")]
+    end
+    C1 -->|"F1: customer message, uploaded evidence"| C2
+    C2 -->|"F2: evidence package and document metadata [B1]"| C3
+    C3 -->|"F3: filtered help document text, provenance flags"| C4
+    C4 -->|"F4: account lookup request"| C5
+    C5 -->|"F5: authenticated read query [B2]"| C6
+    C6 -->|"F6: account attributes, entitlement-scoped results [B2]"| C5
+    C5 -->|"F7: approved account summary"| C4
+    C4 -->|"F8: draft reply content and cited evidence"| C7

images/comparison/E009.mmd

@@ -0,0 +1,16 @@
+flowchart TB
+    subgraph Z_Z1["External Source"]
+        C1["C1: Invoice Inbox"]
+    end
+    subgraph Z_Z2["Guarded Agent Zone"]
+        C2("C2: Invoice Agent")
+        C3[("C3: Draft Review Queue")]
+    end
+    subgraph Z_Z3["Approval and Payments"]
+        C4("C4: Approval Gate")
+        C5("C5: Payment Executor")
+    end
+    C1 -->|"F1: invoice documents and supporting evidence [B1]"| C2
+    C2 -->|"F2: extracted fields and payment draft"| C3
+    C3 -->|"F3: draft package for review [B2]"| C4
+    C4 -->|"F4: approved payment instruction"| C5

images/comparison/E010.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["External sender"]
+        C6["C6: External sender"]
+    end
+    subgraph Z_Z2["Finance operations"]
+        C1("C1: Invoice intake")
+        C2("C2: Invoice agent")
+        C3[("C3: Draft payment record")]
+    end
+    subgraph Z_Z3["Approval and payment execution"]
+        C4("C4: Approval handoff")
+        C5("C5: Payment executor")
+    end
+    C6 -->|"F1: invoice documents, attachments, supporting evidence [B1]"| C1
+    C1 -->|"F2: ingested invoice content and metadata"| C2
+    C2 -->|"F3: extracted payment details, draft invoice fields"| C3
+    C3 -->|"F4: payment draft for approval review [B2]"| C4
+    C4 -->|"F5: approved payment instruction"| C5
+    C5 -->|"F6: payment status, execution result [B2]"| C3

images/comparison/E011.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["External Sources"]
+        C1["C1: Supplier Portal"]
+    end
+    subgraph Z_Z2["Invoice Processing"]
+        C2("C2: Invoice Agent")
+        C3("C3: Document Validation Service")
+        C4[("C4: Draft Ledger")]
+    end
+    subgraph Z_Z3["Approval and Payment"]
+        C5("C5: Approver Console")
+        C6("C6: Payment Executor")
+    end
+    C1 -->|"F1: invoice documents and supporting evidence [B1]"| C2
+    C2 -->|"F2: parsed invoice text, extracted fields, and provenance metadata"| C3
+    C3 -->|"F3: validation result and normalized invoice record"| C2
+    C2 -->|"F4: payment draft and reconciliation notes"| C4
+    C4 -->|"F5: draft payment package for human approval [B2]"| C5
+    C5 -->|"F6: approved payment instruction"| C6

images/comparison/E012.mmd

@@ -0,0 +1,20 @@
+flowchart TB
+    subgraph Z_Z1["Public intake"]
+        C1["C1: Help intake"]
+    end
+    subgraph Z_Z2["Support workspace"]
+        C2("C2: Support agent")
+        C3("C3: Provenance gate")
+        C6("C6: Reply draft")
+    end
+    subgraph Z_Z3["Account services"]
+        C4("C4: Account lookup")
+        C5[("C5: Customer record store")]
+    end
+    C1 -->|"F1: help documents and evidence [B1]"| C3
+    C3 -->|"F2: validated evidence summary and provenance flags"| C2
+    C2 -->|"F3: scoped account query for active ticket [B2]"| C4
+    C4 -->|"F4: read request for allowed customer fields"| C5
+    C5 -->|"F5: ticket-scoped account data"| C4
+    C4 -->|"F6: authorized account result set [B2]"| C2
+    C2 -->|"F7: grounded customer response draft"| C6

images/comparison/E013.mmd

@@ -0,0 +1,19 @@
+flowchart TB
+    subgraph Z_Z1["External Evidence"]
+        C1["C1: Evidence Inbox"]
+    end
+    subgraph Z_Z2["Internal Triage"]
+        C2("C2: Analyst Agent A")
+        C3("C3: Analyst Agent B")
+        C4("C4: Triage Coordinator")
+        C5("C5: Provenance Gate")
+    end
+    subgraph Z_Z3["Case Storage"]
+        C6[("C6: Case Store")]
+    end
+    C1 -->|"F1: incident documents, logs, attachments [B1]"| C2
+    C1 -->|"F2: incident documents, logs, attachments [B1]"| C3
+    C2 -->|"F3: findings, extracted indicators, provenance metadata"| C4
+    C3 -->|"F4: findings, extracted indicators, provenance metadata"| C4
+    C4 -->|"F5: authorized case note draft and write request"| C5
+    C5 -->|"F6: approved case notes, provenance tags, update record [B2]"| C6

images/comparison/E014.mmd

@@ -0,0 +1,17 @@
+flowchart TB
+    subgraph Z_Z1["External Sources"]
+        C4["C4: Help Docs Repository"]
+    end
+    subgraph Z_Z2["Support App"]
+        C1("C1: Support Agent")
+        C2("C2: Response Workspace")
+        C3("C3: Help Document Fetcher")
+    end
+    subgraph Z_Z3["Customer Data"]
+        C5[("C5: Customer Account Store")]
+    end
+    C4 -->|"F1: help articles and support snippets [B1]"| C3
+    C3 -->|"F2: retrieved help content"| C2
+    C5 -->|"F3: customer-account details [B2]"| C2
+    C1 -->|"F4: draft response text and edits"| C2
+    C2 -->|"F5: prepared response and cited facts"| C1

images/comparison/E015.mmd

@@ -0,0 +1,18 @@
+flowchart TB
+    subgraph Z_Z1["External Sources"]
+        C1["C1: Help Document Repository"]
+    end
+    subgraph Z_Z2["Support App Zone"]
+        C2("C2: Support Agent")
+        C3("C3: Response Drafting Service")
+        C4("C4: Policy and Retrieval Filter")
+    end
+    subgraph Z_Z3["Customer Data Zone"]
+        C5[("C5: Customer Account Store")]
+    end
+    C1 -->|"F1: externally supplied help document content and metadata [B1]"| C4
+    C4 -->|"F2: approved excerpts and retrieval results"| C3
+    C2 -->|"F3: support case details and response instructions"| C3
+    C3 -->|"F4: customer-account lookup request [B2]"| C5
+    C5 -->|"F5: customer account attributes and case context [B2]"| C3
+    C3 -->|"F6: draft customer response"| C2