NostromoHub commited on
Commit
8aac1ab
·
verified ·
1 Parent(s): cf18d59

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +156 -3
README.md CHANGED
@@ -1,3 +1,156 @@
1
- ---
2
- license: cc0-1.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc0-1.0
3
+ tags:
4
+ - security
5
+ - cybersecurity
6
+ - cve
7
+ - cisa
8
+ - kev
9
+ - nvd
10
+ - parquet
11
+ - rag
12
+ - public-domain
13
+ pretty_name: CVE-KEV Snapshot (2025-10-29)
14
+ viewer: true
15
+ configs:
16
+ - config_name: preview
17
+ data_files:
18
+ - split: train
19
+ path: parquet/preview.parquet
20
+ - config_name: cve
21
+ data_files:
22
+ - split: train
23
+ path: parquet/cve.parquet
24
+ - config_name: nvd_meta
25
+ data_files:
26
+ - split: train
27
+ path: parquet/nvd_meta.parquet
28
+ - config_name: kev
29
+ data_files:
30
+ - split: train
31
+ path: parquet/kev.parquet
32
+ - config_name: edges
33
+ data_files:
34
+ - split: train
35
+ path: parquet/edges.parquet
36
+ - config_name: rag_meta
37
+ data_files:
38
+ - split: train
39
+ path: rag/meta.parquet
40
+ - config_name: rag_mapping
41
+ data_files:
42
+ - split: train
43
+ path: rag/mapping.parquet
44
+ ---
45
+
46
+ ## CVE-KEV Snapshot (one-time, offline bundle)
47
+
48
+ This bundle lets you rank likely exploited CVEs and cite official sources without any APIs or accounts, fully offline. It’s a one-time snapshot of the last 90 days of NVD, aligned with CISA KEV for immediate focus on likely exploited CVEs. Query-ready Parquet tables and an optional small RAG pack let you rank by severity, pivot by CWE, and fetch references for briefings. Every row includes provenance; validation metrics and an integrity manifest are included. The compiled dataset is dedicated to the public domain (CC0).
49
+
50
+ ## Quick start (60 seconds)
51
+
52
+ - Open `demo/OVX_quickstart.html` (no network) or [demo/OVX_quickstart.ipynb](demo/OVX_quickstart.ipynb).
53
+ - Or run this DuckDB query for top KEV CVEs by severity:
54
+ ```sql
55
+ SELECT c.cve_id, m.cvss_v3_score, k.date_added
56
+ FROM read_parquet('parquet/kev.parquet') k
57
+ JOIN read_parquet('parquet/cve.parquet') c USING (cve_id)
58
+ LEFT JOIN read_parquet('parquet/nvd_meta.parquet') m USING (cve_id)
59
+ ORDER BY (m.cvss_v3_score IS NULL) ASC, m.cvss_v3_score DESC, k.date_added DESC
60
+ LIMIT 20;
61
+ ```
62
+
63
+ ## Snapshot
64
+
65
+ - Snapshot as of (UTC): 2025-09-05T02:12:27Z
66
+ - NVD window (days): 90
67
+
68
+ ## Accuracy at a glance
69
+
70
+ - Total CVE rows: 11,344
71
+ - Total KEV rows: 27
72
+ - Artifacts: [Validation Report](docs/VALIDATION.json) · [Integrity Check](docs/INTEGRITY.txt) · [Build Manifest](docs/BUILD_MANIFEST.json)
73
+ - KEV is window‑aligned; within‑window coverage = 1.0 by design. See `docs/VALIDATION.json` for `kev_rows_total_fetched`, `kev_rows_within_window`, `kev_rows_filtered_out`, and `kev_within_window_over_global_ratio`.
74
+
75
+ ## Contents
76
+
77
+ - `parquet/`: `cve.parquet`, `nvd_meta.parquet`, `kev.parquet`, `edges.parquet`, `preview.parquet` (~1000 CVEs; see preview criteria)
78
+ - `rag/` (optional): `index.faiss`, `meta.parquet`, `mapping.parquet`, `vectors.npy`
79
+ - `docs/`: `LICENSES.md`, `LICENCE.md` (CC0 legal code), `NOTICE.md`, `INTEGRITY.txt`, `VALIDATION.json`, `BUILD_MANIFEST.json`
80
+ - `demo/`: `OVX_quickstart.ipynb`, `OVX_quickstart.html`
81
+
82
+ ## RAG note
83
+
84
+ Optional RAG pack built from PD/CC0 text using `BAAI/bge-small-en-v1.5` (384‑dim), normalized vectors with inner product (cosine) similarity. Retrieval is offline-only; query embedding requires the model to be already cached locally.
85
+
86
+ ## Preview parquet
87
+
88
+ - [parquet/preview.parquet](parquet/preview.parquet) is a convenience subset for quick inspection.
89
+ - Selection: first 1000 rows by `cve_id` ordering from `parquet/cve.parquet`.
90
+ - Columns: `cve_id`, `summary`, `published_date`, `modified_date` (when available); falls back to a best‑effort subset if needed.
91
+
92
+ ## Validation, integrity, build metadata
93
+
94
+ - `docs/VALIDATION.json` includes: counts, `cvss_v3_presence_ratio`, `cwe_presence_ratio`, `kev_cve_coverage_ratio`, `kev_cve_coverage_ratio_within_window`, `kev_rows_total_fetched`, `kev_rows_within_window`, `kev_within_window_over_global_ratio`, `rejected_cve_count`, `url_shape_failures`, `http_head_failures_hard`, `http_head_failures_flaky`, `dead_reference_links`, `duplicate_edges_dropped`, `snapshot_as_of`.
95
+ - URL checks (if present) were executed conservatively with a single worker.
96
+ - Verify integrity locally:
97
+ - macOS: `shasum -a 256 -c docs/INTEGRITY.txt`
98
+ - Linux: `sha256sum -c docs/INTEGRITY.txt`
99
+ - Build metadata (transparency): see `docs/BUILD_MANIFEST.json` for snapshot parameters (timestamp, NVD window, tool/version info, internal commit/config). Provided for transparency; the build system is not included.
100
+
101
+ ## Storage and performance
102
+
103
+ - Parquet compression: `snappy`.
104
+ - Disk footprint varies by window size; see file sizes before downloading.
105
+
106
+ ## Limitations
107
+
108
+ - Some CVEs may lack CVSS vectors/scores within the window (nulls are expected).
109
+ - URL checks are conservative and may still include dead/redirected links; verify with official sources.
110
+ - KEV rows are filtered to the NVD window; counts reflect in‑window coverage, not global totals.
111
+
112
+ ## Usage
113
+
114
+ - DuckDB
115
+ ```sql
116
+ SELECT c.cve_id, m.cvss_v3_score, k.date_added
117
+ FROM read_parquet('parquet/kev.parquet') k
118
+ JOIN read_parquet('parquet/cve.parquet') c USING (cve_id)
119
+ LEFT JOIN read_parquet('parquet/nvd_meta.parquet') m USING (cve_id)
120
+ ORDER BY (m.cvss_v3_score IS NULL) ASC, m.cvss_v3_score DESC, k.date_added DESC
121
+ LIMIT 20;
122
+ ```
123
+
124
+ - Python (Pandas)
125
+ ```python
126
+ import pandas as pd
127
+ k = pd.read_parquet('parquet/kev.parquet')
128
+ c = pd.read_parquet('parquet/cve.parquet')
129
+ m = pd.read_parquet('parquet/nvd_meta.parquet')
130
+ df = k.merge(c, on='cve_id').merge(m[['cve_id','cvss_v3_score']], on='cve_id', how='left')
131
+ print(df.head())
132
+ ```
133
+
134
+ - Python (Polars)
135
+ ```python
136
+ import polars as pl
137
+ k = pl.read_parquet('parquet/kev.parquet')
138
+ c = pl.read_parquet('parquet/cve.parquet')
139
+ m = pl.read_parquet('parquet/nvd_meta.parquet').select('cve_id','cvss_v3_score')
140
+ df = k.join(c, on='cve_id').join(m, on='cve_id', how='left')
141
+ print(df.head())
142
+ ```
143
+
144
+ ## Non-affiliation and licensing
145
+
146
+ - Not affiliated with NIST/NVD, CISA/KEV, or FIRST/EPSS.
147
+ - NVD non-endorsement: "This product uses data from the NVD API but is not endorsed or certified by the NVD."
148
+ - Compiled artifact dedicated to the public domain under CC0 1.0 (see `docs/LICENCE.md`).
149
+ - Upstream sources: NVD (public domain), CISA KEV (CC0). Linked third‑party pages (vendor advisories, blogs, etc.) are governed by their own terms.
150
+
151
+ ## Citation
152
+
153
+ If you use this snapshot, please cite:
154
+ ```
155
+ "CVE-KEV Snapshot (${snapshot_as_of})", CC0-1.0, https://huggingface.co/datasets/NostromoHub/cve-kev-snapshot-90d-2025-10-29
156
+ ```