Update merged CVE-CWE dataset

.gitattributes

@@ -59,3 +59,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 Global_Dataset.csv filter=lfs diff=lfs merge=lfs -text
+CVE_CWE_Dataset.csv filter=lfs diff=lfs merge=lfs -text

CVE_CWE_Dataset.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:243af33535d71ddd784ec06d1f286f2fede4075586afa6f4d139c370f56a71cc
+size 128539315

README.md

@@ -5,6 +5,7 @@ tags:
 - cybersecurity
 - vulnerability
 - cve
+- nvd
 - cwe
 - software-security
 task_categories:
@@ -17,26 +18,58 @@ license: apache-2.0
 
 ## Dataset description
 
-This dataset maps Common Vulnerabilities and Exposures (CVEs) to Common Weakness Enumeration (CWE) entries in the CWE-699 Software category. It combines CVE descriptions with CWE descriptions and parent-category information for security research and vulnerability classification.
+This dataset merges two CVE-to-CWE data sources into a single table:
+
+- The original CVE-to-CWE mapping dataset from
+  [zefang-liu/cve-and-cwe-mapping-dataset](https://huggingface.co/datasets/zefang-liu/cve-and-cwe-mapping-dataset),
+  enriched with CWE descriptions and parent-category information from the scraped CWE-699 software hierarchy.
+- The NVD CVE records scraped by this project, resolved against the CWE-699 software hierarchy.
+
+The two sources are joined on the CVE identifier. Variables shared by both sources are matched onto a single
+column; variables that exist in only one source are kept as separate columns and left empty for rows coming
+from the other source.
 
 ## Dataset structure
 
-The dataset is provided as `Global_Dataset.csv`. Its main fields include:
+The dataset is provided as `CVE_CWE_Dataset.csv`.
+
+Variables shared by both sources:
+
+- `CVE-ID`: CVE identifier (join key).
+- `YEAR`: year the CVE was published.
+- `CVE-DESCRIPTION`: vulnerability description.
+- `CVE-DESCRIPTION-NORMALIZED`: normalized vulnerability description.
+- `CWE-ID`: associated CWE identifier(s), pipe separated.
+
+Variables from the original (zefang-liu) dataset only:
+
+- `CWE-DESCRIPTION`: description of the associated weakness.
+- `PARENTS`: parent CWE category or categories.
+- `IS-IN-CWE699`: whether the CWE belongs to the CWE-699 Software category.
+- `CVSS-V3`: CVSS v3 score.
+- `CVSS-V2`: CVSS v2 score.
+- `SEVERITY`: severity rating.
+- `ID`: identifier from the original dataset.
+
+Variables from the NVD dataset only:
 
-- `CVE-ID`: CVE identifier.
-- `CVE-DESCRIPTION`: Description of the vulnerability.
-- `CWE-ID`: Associated CWE identifier.
-- `CWE-DESCRIPTION`: Description of the associated weakness.
-- `PARENTS`: Parent CWE category or categories.
-- `YEAR`: Year extracted from the CVE identifier.
-- `IS-IN-CWE699`: Whether the CWE belongs to the CWE-699 Software category.
+- `published`: publication timestamp.
+- `last_modified`: last modification timestamp.
+- `all_nvd_cwes`: all CWE identifiers NVD associates with the CVE (pipe separated).
+- `number_of_nvd_cwes`: count of `all_nvd_cwes`.
+- `number_of_cwe_699_matches`: count of `CWE-ID`.
+- `true_children`: the CWE-699 child ids from `CWE-ID` that exist in `CWE_SOFTWARE.json` (list).
+- `true_categories`: the CWE-699 category ids those children belong to (sorted list).
 
 ## Data sources
 
-- CVE-to-CWE mappings and vulnerability descriptions are sourced from the `zefang-liu/cve-and-cwe-mapping-dataset` dataset on Hugging Face.
-- CWE descriptions and hierarchy information are collected from the CWE software-category data scraped by this project.
+- CVE-to-CWE mappings and vulnerability descriptions are sourced from the `zefang-liu/cve-and-cwe-mapping-dataset`
+  dataset on Hugging Face.
+- Raw CVE records are collected from the NVD.
+- CWE descriptions, hierarchy and CWE-699 Software category membership are collected from the CWE data scraped
+  by this project.
 
-The dataset is generated by `dataset/generate_dataset.ipynb`.
+The dataset is generated by `dataset/merge_datasets.ipynb`.
 
 ## Intended uses
 
@@ -49,8 +82,11 @@ This dataset is intended for research and experimentation involving:
 
 ## Limitations
 
-The mappings reflect the upstream data sources and may contain incomplete, ambiguous, or changing vulnerability information. The dataset should not be treated as a definitive security assessment or as a replacement for reviewing the original NVD and CWE records.
+The mappings reflect the upstream NVD, CVE and CWE data and may contain incomplete, ambiguous, or changing
+vulnerability information. The dataset should not be treated as a definitive security assessment or as a
+replacement for reviewing the original NVD and CWE records.
 
 ## License and attribution
 
-Please review and follow the licenses and attribution requirements of the upstream CVE and CWE sources before redistributing or using this dataset.
+Please review and follow the licenses and attribution requirements of the upstream NVD, CVE and CWE sources
+before redistributing or using this dataset.