Add dataset card
Browse files
README.md
ADDED
@@ -0,0 +1,198 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
# YAML Metadata Block
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
tags:
|
6 |
+
- vulnerability-detection
|
7 |
+
- cve
|
8 |
+
- code-changes
|
9 |
+
- software-security
|
10 |
+
- stratified-split
|
11 |
+
license: mit
|
12 |
+
dataset_info:
|
13 |
+
features: # Features in the *final split files*
|
14 |
+
- name: idx
|
15 |
+
dtype: int64
|
16 |
+
- name: func_before
|
17 |
+
dtype: string
|
18 |
+
- name: Vulnerability Classification
|
19 |
+
dtype: string
|
20 |
+
- name: vul
|
21 |
+
dtype: int64
|
22 |
+
- name: func_after
|
23 |
+
dtype: string
|
24 |
+
- name: patch
|
25 |
+
dtype: string
|
26 |
+
- name: CWE ID
|
27 |
+
dtype: string
|
28 |
+
- name: lines_before
|
29 |
+
dtype: string
|
30 |
+
- name: lines_after
|
31 |
+
dtype: string
|
32 |
+
splits:
|
33 |
+
- name: train
|
34 |
+
num_examples: 150909
|
35 |
+
- name: validation
|
36 |
+
num_examples: 18864
|
37 |
+
- name: test
|
38 |
+
num_examples: 18863
|
39 |
+
|
40 |
+
|
41 |
+
dataset_original_file_size: 10GB uuncompressed
|
42 |
+
---
|
43 |
+
|
44 |
+
# MSR Data Cleaned - C/C++ Code Vulnerability Dataset
|
45 |
+
|
46 |
+
[](LICENSE)
|
47 |
+
|
48 |
+
|
49 |
+
|
50 |
+
## π Dataset Description
|
51 |
+
A curated collection of C/C++ code vulnerabilities paired with:
|
52 |
+
- CVE details (scores, classifications, exploit status)
|
53 |
+
- Code changes (commit messages, added/deleted lines)
|
54 |
+
- File-level and function-level diffs
|
55 |
+
|
56 |
+
## π Sample Data Structure from original file
|
57 |
+
```python
|
58 |
+
+---------------+-----------------+----------------------+---------------------------+
|
59 |
+
| CVE ID | Attack Origin | Publish Date | Summary |
|
60 |
+
+===============+=================+======================+===========================+
|
61 |
+
| CVE-2015-8467 | Remote | 2015-12-29 | "The samldb_check_user..."|
|
62 |
+
+---------------+-----------------+----------------------+---------------------------+
|
63 |
+
| CVE-2016-1234 | Local | 2016-01-15 | "Buffer overflow in..." |
|
64 |
+
+---------------+-----------------+----------------------+---------------------------+
|
65 |
+
|
66 |
+
```
|
67 |
+
Note: This is a simplified preview; the full dataset includes additional fields like commit_id, func_before, etc.
|
68 |
+
|
69 |
+
|
70 |
+
### 1. Accessing in Colab
|
71 |
+
```python
|
72 |
+
!pip install huggingface_hub -q
|
73 |
+
from huggingface_hub import snapshot_download
|
74 |
+
|
75 |
+
repo_id = "starsofchance/MSR_data_cleaned"
|
76 |
+
dataset_path = snapshot_download(repo_id=repo_id, repo_type="dataset")
|
77 |
+
```
|
78 |
+
|
79 |
+
### 2. Extracting the Dataset
|
80 |
+
```python
|
81 |
+
!apt-get install unzip -qq
|
82 |
+
!unzip "/root/.cache/huggingface/.../MSR_data_cleaned.zip" -d "/content/extracted_data"
|
83 |
+
```
|
84 |
+
**Note: Extracted size is 10GB (1.5GB compressed). Ensure sufficient disk space.
|
85 |
+
|
86 |
+
### 3. Creating Splits (Colab Pro Recommended)
|
87 |
+
We used this memory-efficient approach:
|
88 |
+
```python
|
89 |
+
from datasets import load_dataset
|
90 |
+
dataset = load_dataset("csv", data_files="MSR_data_cleaned.csv", streaming=True)
|
91 |
+
|
92 |
+
# Randomly distribute rows (80-10-10)
|
93 |
+
for row in dataset:
|
94 |
+
rand = random.random()
|
95 |
+
if rand < 0.8: write_to(train.csv)
|
96 |
+
elif rand < 0.9: write_to(validation.csv)
|
97 |
+
else: write_to(test.csv)
|
98 |
+
```
|
99 |
+
|
100 |
+
|
101 |
+
**Hardware Requirements:**
|
102 |
+
- Minimum 25GB RAM
|
103 |
+
- Strong CPU (Colab Pro T4 GPU recommended)
|
104 |
+
|
105 |
+
##π Dataset Statistics
|
106 |
+
|
107 |
+
- Number of Rows: 188,636
|
108 |
+
- Vulnerability Distribution:
|
109 |
+
- Vulnerable (1): 18,863 (~10%)
|
110 |
+
- Non-Vulnerable (0): 169,773 (~90%)
|
111 |
+
##π Data Fields Description
|
112 |
+
- CVE_ID: Unique identifier for the vulnerability (Common Vulnerabilities and Exposures).
|
113 |
+
- CWE_ID: Weakness category identifier (Common Weakness Enumeration).
|
114 |
+
- Score: CVSS score indicating severity (float, 0-10).
|
115 |
+
- Summary: Brief description of the vulnerability.
|
116 |
+
- commit_id: Git commit hash linked to the code change.
|
117 |
+
- codeLink: URL to the code repository or commit.
|
118 |
+
- file_name: Name of the file containing the vulnerability.
|
119 |
+
- func_after: Function code after the change.
|
120 |
+
- lines_after: Code lines after the change.
|
121 |
+
- Access_Gained: Type of access gained by exploiting the vulnerability.
|
122 |
+
- Attack_Origin: Source of the attack (e.g., Remote, Local).
|
123 |
+
- Authentication_Required: Whether authentication is needed to exploit.
|
124 |
+
- Availability: Impact on system availability.
|
125 |
+
- CVE_Page: URL to the CVE details page.
|
126 |
+
- Complexity: Complexity of exploiting the vulnerability.
|
127 |
+
- Confidentiality: Impact on data confidentiality.
|
128 |
+
- Integrity: Impact on data integrity.
|
129 |
+
- Known_Exploits: Details of known exploits, if any.
|
130 |
+
- Publish_Date: Date the vulnerability was published.
|
131 |
+
- Update_Date: Date of the last update to the vulnerability data.
|
132 |
+
- Vulnerability_Classification: Type or category of the vulnerability.
|
133 |
+
- add_lines: Lines added in the commit.
|
134 |
+
- del_lines: Lines deleted in the commit.
|
135 |
+
- commit_message: Description of the commit.
|
136 |
+
- files_changed: List of files modified in the commit.
|
137 |
+
- func_before: Function code before the change.
|
138 |
+
- lang: Programming language (e.g., C, C++).
|
139 |
+
- lines_before: Code lines before the change.
|
140 |
+
|
141 |
+
|
142 |
+
## splits file for UltiVul project:
|
143 |
+
|
144 |
+
## π Sample Data Structure (from train.csv)
|
145 |
+
```python
|
146 |
+
{
|
147 |
+
'idx': 0, # Unique ID within the train split
|
148 |
+
'func_before': '...', # String containing function code before change
|
149 |
+
'Vulnerability Classification': '...', # Original vulnerability type classification
|
150 |
+
'vul': 0, # Integer: 0 for non-vulnerable, 1 for vulnerable (target label)
|
151 |
+
'func_after': '...', # String containing function code after change
|
152 |
+
'patch': '...', # String containing diff patch
|
153 |
+
'CWE ID': '...', # String CWE ID, e.g., "CWE-119"
|
154 |
+
'lines_before': '...', # String lines before change context
|
155 |
+
'lines_after': '...' # String lines after change context
|
156 |
+
}
|
157 |
+
```
|
158 |
+
**Note: This shows the structure of the final split files (train.csv, validation.csv, test.csv). The original MSR_data_cleaned.csv contains many more metadata fields.
|
159 |
+
|
160 |
+
|
161 |
+
##π¦ Dataset New Files
|
162 |
+
The dataset is available as three CSV files (specially created for the UltiVul project) hosted on Hugging Face, uploaded via huggingface_hub:
|
163 |
+
|
164 |
+
- train.csv
|
165 |
+
Size: 667 MB
|
166 |
+
Description: Training split with 150,909 samples, approximately 80% of the data.
|
167 |
+
- validation.csv
|
168 |
+
Size: 86 MB
|
169 |
+
Description: Validation split with 18,864 samples, approximately 10% of the data.
|
170 |
+
- test.csv
|
171 |
+
Size: 84.8 MB
|
172 |
+
Description: Test split with 18,863 samples, approximately 10% of the data.
|
173 |
+
|
174 |
+
|
175 |
+
|
176 |
+
π Acknowledgements
|
177 |
+
Original dataset provided by Fan et al., 2020
|
178 |
+
Thanks to the Hugging Face team for dataset hosting tools.
|
179 |
+
|
180 |
+
## π Citation
|
181 |
+
```bibtex
|
182 |
+
@inproceedings{fan2020ccode,
|
183 |
+
title={A C/C++ Code Vulnerability Dataset with Code Changes and CVE Summaries},
|
184 |
+
author={Fan, Jiahao and Li, Yi and Wang, Shaohua and Nguyen, Tien N},
|
185 |
+
booktitle={MSR '20: 17th International Conference on Mining Software Repositories},
|
186 |
+
pages={1--5},
|
187 |
+
year={2020},
|
188 |
+
doi={10.1145/3379597.3387501}
|
189 |
+
}
|
190 |
+
```
|
191 |
+
|
192 |
+
## π Dataset Creation
|
193 |
+
- **Source**: Original data from [MSR 2020 Paper](https://doi.org/10.1145/3379597.3387501)
|
194 |
+
- **Processing**:
|
195 |
+
- Cleaned and standardized CSV format
|
196 |
+
- Stream-based splitting to handle large size
|
197 |
+
- Preserved all original metadata
|
198 |
+
|