Generate datasets amd models based on vulnerabilities descriptions from Vulnerability-Lookup.
Project description
VulnTrain
Generate datasets amd models based on vulnerabilities descriptions from Vulnerability-Lookup.
Uses data from the vulnerability-lookup:meta container such as vulnrichment and FKIE.
Datasets
Various datasets generated are available on HuggingFace:
https://huggingface.co/datasets/circl/vulnerability
Usage
Generate datasets
Authenticate to HuggingFace:
huggingface-cli login
Install VulnTrain:
$ pipx install VulnTrain
Then ensures that the kvrocks database of Vulnerability-Lookup is running.
Creation of datasets:
$ vulntrain-create-dataset --nb-rows 10000 --upload --repo-id CIRCL/vulnerability-dataset-10k
Generating train split: 9999 examples [00:00, 177710.74 examples/s]
DatasetDict({
train: Dataset({
features: ['id', 'title', 'description', 'cpes'],
num_rows: 8999
})
test: Dataset({
features: ['id', 'title', 'description', 'cpes'],
num_rows: 1000
})
})
Creating parquet from Arrow format: 100%|██████████████████████████████████████████████████████████████████████████████| 9/9 [00:00<00:00, 49.66ba/s]
Uploading the dataset shards: 100%|████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:02<00:00, 2.03s/it]
Creating parquet from Arrow format: 100%|██████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 63.36ba/s]
Uploading the dataset shards: 100%|████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:01<00:00, 1.19s/it]
README.md: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████| 503/503 [00:00<00:00, 2.34MB/s]
Train
Training for text generation
For now we are using distilbert-base-uncased (AutoModelForMaskedLM) or gpt2 (AutoModelForCausalLM). The goal is to generate text.
$ vulntrain-train-dataset --base-model gpt2 --model-name CIRCL/vulnerability
Using CUDA (Nvidia GPU).
[codecarbon WARNING @ 13:28:13] Multiple instances of codecarbon are allowed to run at the same time.
[codecarbon INFO @ 13:28:13] [setup] RAM Tracking...
[codecarbon INFO @ 13:28:13] [setup] CPU Tracking...
[codecarbon WARNING @ 13:28:13] No CPU tracking mode found. Falling back on CPU constant mode.
Linux OS detected: Please ensure RAPL files exist at \sys\class\powercap\intel-rapl to measure CPU
[codecarbon WARNING @ 13:28:14] We saw that you have a AMD EPYC 9124 16-Core Processor but we don't know it. Please contact us.
[codecarbon INFO @ 13:28:14] CPU Model on constant consumption mode: AMD EPYC 9124 16-Core Processor
[codecarbon INFO @ 13:28:14] [setup] GPU Tracking...
[codecarbon INFO @ 13:28:14] Tracking Nvidia GPU via pynvml
[codecarbon INFO @ 13:28:14] >>> Tracker's metadata:
[codecarbon INFO @ 13:28:14] Platform system: Linux-6.8.0-48-generic-x86_64-with-glibc2.39
[codecarbon INFO @ 13:28:14] Python version: 3.12.3
[codecarbon INFO @ 13:28:14] CodeCarbon version: 2.8.3
[codecarbon INFO @ 13:28:14] Available RAM : 251.586 GB
[codecarbon INFO @ 13:28:14] CPU count: 64
[codecarbon INFO @ 13:28:14] CPU model: AMD EPYC 9124 16-Core Processor
[codecarbon INFO @ 13:28:14] GPU count: 2
[codecarbon INFO @ 13:28:14] GPU model: 2 x NVIDIA L40S
[codecarbon INFO @ 13:28:18] Saving emissions data to file /home/cedric/VulnTrain/emissions.csv | 1/2700 [00:07<5:45:36, 7.68s/it]
...
...
...
Training for classification
tf-idf on the vulnerability descriptions.
Validation
It is possible to send prompts to a model trained for text generation (descriptions of vulnerabilities).
$ vulntrain-validate-text-generation --help
usage: vulntrain-validate-text-generation [-h] [--model MODEL] [--prompt PROMPT]
Validate a text generation model for vulnerabilities.
options:
-h, --help show this help message and exit
--model MODEL The model to use.
--prompt PROMPT The prompt for the generator.
Example:
$ vulntrain-validate-text-generation --prompt "A new vulnerability in OpenSSL allows attackers to" --model CIRCL/vulnerability
config.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 907/907 [00:00<00:00, 6.70MB/s]
model.safetensors: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 498M/498M [00:12<00:00, 41.3MB/s]
generation_config.json: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 119/119 [00:00<00:00, 1.63MB/s]
tokenizer_config.json: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 556/556 [00:00<00:00, 4.01MB/s]
vocab.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 798k/798k [00:00<00:00, 3.25MB/s]
merges.txt: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 456k/456k [00:00<00:00, 5.58MB/s]
tokenizer.json: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3.56M/3.56M [00:00<00:00, 10.3MB/s]
special_tokens_map.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 470/470 [00:00<00:00, 3.51MB/s]
Device set to use cuda:0
Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation=True` to explicitly truncate examples to max length. Defaulting to 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to `truncation`.
[{'generated_text': 'A new vulnerability in OpenSSL allows attackers to cause a Denial of Service (DoS) when receiving a specially crafted SIP message.\n\n\nThis issue affects: OpenSSL versions prior to 1.2.1\n\n\n\n * OpenSSL 1.2.1 prior to 1.2.1-HF1, which fixes this issue.\n\n * OpenSSL version 1.2.1 prior to 1.2.1-HF1 and OpenSSL 1.2.2 prior'}]
License
VulnTrain is licensed under GNU General Public License version 3
Copyright (c) 2025 Computer Incident Response Center Luxembourg (CIRCL)
Copyright (C) 2025 Cédric Bonhomme - https://github.com/cedricbonhomme
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file vulntrain-0.4.0.tar.gz.
File metadata
- Download URL: vulntrain-0.4.0.tar.gz
- Upload date:
- Size: 8.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.12.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
35f0270c83b2e86bb390fc4bc9c4fd7a282df2549d7edda072ebb8babdbbabad
|
|
| MD5 |
16d8950776e4bd35240ca3e37481a3b7
|
|
| BLAKE2b-256 |
f41b8bfecb4fedd8b638144c9408d2eb8c5e07e937f4b18d7b2a4c3d23552033
|
Provenance
The following attestation bundles were made for vulntrain-0.4.0.tar.gz:
Publisher:
release.yml on vulnerability-lookup/VulnTrain
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
vulntrain-0.4.0.tar.gz -
Subject digest:
35f0270c83b2e86bb390fc4bc9c4fd7a282df2549d7edda072ebb8babdbbabad - Sigstore transparency entry: 173329908
- Sigstore integration time:
-
Permalink:
vulnerability-lookup/VulnTrain@d922d3a053a7fb78773b7ecc4b9de691e419be08 -
Branch / Tag:
refs/tags/v0.4.0 - Owner: https://github.com/vulnerability-lookup
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
release.yml@d922d3a053a7fb78773b7ecc4b9de691e419be08 -
Trigger Event:
release
-
Statement type:
File details
Details for the file vulntrain-0.4.0-py3-none-any.whl.
File metadata
- Download URL: vulntrain-0.4.0-py3-none-any.whl
- Upload date:
- Size: 9.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.12.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
93ff9a44d9d92fe9de24eaaa364ac43ca0afe83fdf7b574d9fe14fedfe6737c5
|
|
| MD5 |
f41cf3e127ec6e69a7b4daa0f5381160
|
|
| BLAKE2b-256 |
94f8ef7daade6c332de0d5f410fd9eb7220e669532b9cddcaa1a63b38b634530
|
Provenance
The following attestation bundles were made for vulntrain-0.4.0-py3-none-any.whl:
Publisher:
release.yml on vulnerability-lookup/VulnTrain
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
vulntrain-0.4.0-py3-none-any.whl -
Subject digest:
93ff9a44d9d92fe9de24eaaa364ac43ca0afe83fdf7b574d9fe14fedfe6737c5 - Sigstore transparency entry: 173329911
- Sigstore integration time:
-
Permalink:
vulnerability-lookup/VulnTrain@d922d3a053a7fb78773b7ecc4b9de691e419be08 -
Branch / Tag:
refs/tags/v0.4.0 - Owner: https://github.com/vulnerability-lookup
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
release.yml@d922d3a053a7fb78773b7ecc4b9de691e419be08 -
Trigger Event:
release
-
Statement type: