Datasets:
The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
index: int64
Access Gained: string
Attack Origin: string
Authentication Required: string
Availability: string
CVE ID: string
CVE Page: string
CWE ID: string
Complexity: string
Confidentiality: string
Integrity: string
Known Exploits: double
Publish Date: string
Score: double
Summary: string
Update Date: string
Vulnerability Classification: string
add_lines: int64
codeLink: string
commit_id: string
commit_message: string
del_lines: int64
file_name: string
files_changed: string
func_after: string
func_before: string
lang: string
lines_after: string
lines_before: string
parentID: string
patch: string
project: string
project_after: string
project_before: string
target: int64
vul_func_with_fix: string
processed_func: string
flaw_line: string
flaw_line_index: string
before_fix: null
after_fix: null
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 5001
to
{'before_fix': Value('string'), 'after_fix': Value('string'), 'target': Value('int64')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
return get_rows(
^^^^^^^^^
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2431, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1952, in __iter__
for key, pa_table in self._iter_arrow():
^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1984, in _iter_arrow
pa_table = cast_table_to_features(pa_table, self.features)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2192, in cast_table_to_features
raise CastError(
datasets.table.CastError: Couldn't cast
index: int64
Access Gained: string
Attack Origin: string
Authentication Required: string
Availability: string
CVE ID: string
CVE Page: string
CWE ID: string
Complexity: string
Confidentiality: string
Integrity: string
Known Exploits: double
Publish Date: string
Score: double
Summary: string
Update Date: string
Vulnerability Classification: string
add_lines: int64
codeLink: string
commit_id: string
commit_message: string
del_lines: int64
file_name: string
files_changed: string
func_after: string
func_before: string
lang: string
lines_after: string
lines_before: string
parentID: string
patch: string
project: string
project_after: string
project_before: string
target: int64
vul_func_with_fix: string
processed_func: string
flaw_line: string
flaw_line_index: string
before_fix: null
after_fix: null
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 5001
to
{'before_fix': Value('string'), 'after_fix': Value('string'), 'target': Value('int64')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Dataset Card for Linevul Dataset (2008-2026)
Note
- file to use: my_dataset_clean.csv
Dataset Details
Dataset Description
This dataset consists of C++ code functions curated for software vulnerability detection. It is designed to fine-tune RoBERTa-based models into LineVul models, which identify security vulnerabilities at a function or line level. The data covers a span from 2008 to 2026, capturing diverse C++ coding patterns and historical security flaws.
- Curated by: Myself
- Funded by: Self-funded for Academic Research
- Language(s) (NLP): C++ (Source Code)
- License: MIT (Methodology based on Fan et al., 2020)
Dataset Sources
- Repository: Github
- Paper: A C/C++ Code Vulnerability Dataset with Code Changes and CVE Summaries
Uses
Direct Use
The primary use case is fine-tuning Large Language Models (LLMs) like RoBERTa to classify code snippets as "vulnerable" or "benign." It is also useful for differential analysis between patched and unpatched code.
Out-of-Scope Use
This dataset is not intended for the automated generation of malicious exploits or for use in unauthorized security testing.
Dataset Structure
The dataset is provided in a single CSV file containing approximately 104,000 examples.
| Field | Description |
|---|---|
processed_func |
The raw C++ function containing the vulnerability. |
vul_func_with_fix |
The patched C++ function with the security fix applied. |
target |
Binary label: 1 (Vulnerable/Before Fix) or 0 (Benign/After Fix). |
Recommended Split
- Training: 80% (~83,200 samples)
- Testing/Validation: 20% (~20,800 samples)
Dataset Creation
Curation Rationale
Developed as part of a Master's degree project to provide a high-quality, long-term dataset (2008-2026) for training state-of-the-art vulnerability detection models.
Source Data
Data Collection and Processing
The data was crawled following the methodology of the BigVul paper. It targets GitHub commits linked to Common Vulnerabilities and Exposures (CVE) entries.
- Extraction: Functions were extracted from security-relevant commits.
- Labeling: Snippets before the fix are labeled
1, and snippets after the fix are labeled0. - Cleaning: Removal of non-C++ code and duplicate entries.
Who are the source data producers?
The data originates from open-source developers contributing to public C++ repositories and security researchers documenting CVEs.
Bias, Risks, and Limitations
- CVE-Centric: The dataset only includes vulnerabilities that were officially caught and patched; it may not represent undiscovered logic flaws.
- Label Leakage: Because
target 0is often theafter_fixversion oftarget 1, models may learn to recognize "fix patterns" (like adding a null check) rather than the vulnerability itself.
- Downloads last month
- 15