datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
securecode-web
SecureCode Web: Traditional Web & Application Security Dataset
Production-grade web security vulnerability dataset with complete incident grounding, 4-turn conversational structure, and comprehensive operational guidance
Paper | GitHub | Dataset | Model Collection | Blog Post
What's new in v2.6
v2.6 restores proper Express.js coverage for the topics whose examples were removed in v2.5.1 (they had
shared one reused answer). 29 new, genuinely distinct Express.js… See the full description on the dataset page: https://huggingface.co/datasets/scthornton/securecode-web.SecureVibeBench
SecureVibeBench: First Secure Vibe Coding Benchmark
SecureVibeBench is a benchmark consisting of 105 C/C++ secure coding tasks sourced from 41 projects in OSS-Fuzz for code agents. It is designed to evaluate secure vibe coding by reconstructing real-world scenarios where human developers introduced vulnerabilities.
Paper: SecureVibeBench: Benchmarking Secure Vibe Coding of AI Agents via Reconstructing Vulnerability-Introducing Scenarios
Repository: iCSawyer/SecureVibeBench
Venue:… See the full description on the dataset page: https://huggingface.co/datasets/iCSawyer/SecureVibeBench.adv_glue
Dataset Card for Adversarial GLUE
Dataset Summary
Adversarial GLUE Benchmark (AdvGLUE) is a comprehensive robustness evaluation benchmark that focuses on the adversarial robustness evaluation of language models. It covers five natural language understanding tasks from the famous GLUE tasks and is an adversarial version of GLUE benchmark.
AdvGLUE considers textual adversarial attacks from different perspectives and hierarchies, including word-level transformations… See the full description on the dataset page: https://huggingface.co/datasets/AI-Secure/adv_glue.SecureCodeV2_qwen-2.5-7b-instruct_tokenized_vulnerablesecurecode
SecureCode: Comprehensive Security Training Dataset for AI Coding Assistants
The largest open security training dataset for AI coding assistants, covering both traditional web security and AI/ML security
Overview
SecureCode combines 2,372 security-focused training examples into a single, unified dataset with HuggingFace configs for flexible loading. Every example provides vulnerable code, explains why it's dangerous, demonstrates a secure alternative, and… See the full description on the dataset page: https://huggingface.co/datasets/scthornton/securecode.securecode-web-archive
SecureCode Web: Traditional Web & Application Security Dataset
Production-grade web security vulnerability dataset with complete incident grounding, 4-turn conversational structure, and comprehensive operational guidance
Paper | GitHub | Dataset | Model Collection | Blog Post
What's new in v2.6
v2.6 restores proper Express.js coverage for the topics whose examples were removed in v2.5.1 (they had
shared one reused answer). 29 new, genuinely distinct Express.js… See the full description on the dataset page: https://huggingface.co/datasets/ChipHolmes/securecode-web-archive.securecode-aiml
SecureCode AI/ML: AI/ML Security Training Dataset
Security training dataset for AI coding assistants covering the OWASP LLM Top 10 2025 categories. 747 examples of AI/ML vulnerabilities with vulnerable code, secure implementations, and defense-in-depth guidance.
Grounding note (audited 2026-07-06). Every context.real_world_example was independently
fact-checked. The vulnerability classes and the CVEs cited in the conversations (e.g. EchoLeak
CVE-2025-32711, EmailGPT… See the full description on the dataset page: https://huggingface.co/datasets/scthornton/securecode-aiml.http-attack-requests
HTTP Attack Requests — multi-class
Real HTTP requests labelled with the web-attack class carried in the request,
for training and evaluating request/payload classifiers (WAF / DAST style).
Classes (7)
normal, sqli, xss, ssrf, ssti, lfi, traversal
(IDOR is intentionally excluded — it's an access-control flaw with no payload signature.)
How it was built (and why it's shortcut-resistant)
Every example is a full HTTP request built on the same real… See the full description on the dataset page: https://huggingface.co/datasets/SecureAI-SE/http-attack-requests.SecureFinAI_Contest_2025-Task_3_SpanishOCRfuzz-agent-trajectories
Agentic Web-API Fuzzing Trajectories
Verified multi-turn tool-calling trajectories of an LLM agent fuzzing a deliberately-vulnerable
web API for eight vulnerability classes (sqli, bola, bfla, mass-assignment, command-injection,
excessive-data-exposure, error-500, verbose-errors).
Each trajectory was produced by a strong teacher running the think -> act (http_request) ->
observe loop, and kept only if its reports were verified against ground truth (precise +
covering enough… See the full description on the dataset page: https://huggingface.co/datasets/SecureAI-SE/fuzz-agent-trajectories.Benchmarks_CyberSec_SECURE
Dataset Card for SECURE (RISys-Lab Mirror)
⚠️ Disclaimer: > This repository is a mirror/re-host of the original SECURE benchmark.RISys-Lab is not the author of this dataset. We are hosting this copy in Parquet format to ensure seamless integration and stability for our internal evaluation pipelines. All credit belongs to the original authors listed below.
Repository Intent
This Hugging Face dataset is a re-host of the original SECURE benchmark. It has been converted… See the full description on the dataset page: https://huggingface.co/datasets/RISys-Lab/Benchmarks_CyberSec_SECURE.Eve-Secure-Coder
Eve Secure Coder Dataset
Dataset Description
Eve Secure Coder is a composite dataset curated to fine-tune Qwen3-4b-Heretic into a highly capable, security-conscious coding assistant with a distinct personality and no refusals.
The primary goal of this dataset is to bridge the gap between sterile, robotic coding assistants and engaging, conversational AI, without sacrificing technical accuracy or security. By combining high-quality coding archives, vulnerability detection… See the full description on the dataset page: https://huggingface.co/datasets/TitleOS/Eve-Secure-Coder.securecode-archive
SecureCode: Comprehensive Security Training Dataset for AI Coding Assistants
The largest open security training dataset for AI coding assistants, covering both traditional web security and AI/ML security
Overview
SecureCode combines 2,372 security-focused training examples into a single, unified dataset with HuggingFace configs for flexible loading. Every example provides vulnerable code, explains why it's dangerous, demonstrates a secure alternative, and… See the full description on the dataset page: https://huggingface.co/datasets/ChipHolmes/securecode-archive.securecode-aiml-archive
SecureCode AI/ML: AI/ML Security Training Dataset
Security training dataset for AI coding assistants covering the OWASP LLM Top 10 2025 categories. 747 examples of AI/ML vulnerabilities with vulnerable code, secure implementations, and defense-in-depth guidance.
Grounding note (audited 2026-07-06). Every context.real_world_example was independently
fact-checked. The vulnerability classes and the CVEs cited in the conversations (e.g. EchoLeak
CVE-2025-32711, EmailGPT… See the full description on the dataset page: https://huggingface.co/datasets/ChipHolmes/securecode-aiml-archive.SecureFinAI_Contest_2025-Task_3_EnglishOCRprogressively-more-secure-system-prompt
What it is
This dataset takes the provided 'secure' system prompt and breaks it down into (human-annotated) atomic chunks that add constraints.
The combinations and their products are then reconstructed into subsets of the original system prompt, for iterative checking.
What it's for
To see at which point a model using this system prompt can be sent on or off task
Number of Chunks
intent [3]
capbilities [3]
policy [12]
examples [3]
terminator [1]… See the full description on the dataset page: https://huggingface.co/datasets/Mindgard/progressively-more-secure-system-prompt.securecode-vulnerablerkl-secure-reasoning-brief-telemetry
RKL Secure Reasoning Brief - Research Telemetry Dataset
⚠️ Development Transparency: This project was developed with extensive AI coding assistance (Claude Code, ChatGPT) under tight time constraints for the Kaggle AI Agents Capstone. The developer designed the architecture, telemetry schema, and system integration decisions; AI tools scaffolded most of the implementation code. This is an honest exploratory prototype built to learn what telemetry is valuable to collect. Full… See the full description on the dataset page: https://huggingface.co/datasets/rkl-org/rkl-secure-reasoning-brief-telemetry.labelled-secure_code_dataset
labelled-secure_code_dataset
Dataset Description
XLAMv2 / APIGen Tool Calling dataset
Dataset Details
Created by: Always Further
License: CC BY 4.0
Language(s): [English
Dataset Size: 10050
Data Splits
[train]
Dataset Creation
This dataset was created using DeepFabric, an open-source tool for generating high-quality training datasets for AI models.
Usage
from datasets import load_dataset
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/nolabs/labelled-secure_code_dataset.SECURE-VOODopencode-apihack-trajectoriessecure_1_with_code_datasetSECURE-CWETmalware-text-db-securebert-ner-512SECURE-MAETasia-owid-number-of-transboundary-animal-breeds-which-have-genetic-resources-secured-in-conserva
Number Of Transboundary Animal Breeds Which Have Genetic Resources Secured In Conservation Facilities | Asia (Our World in Data)
🌏 203 observations · 24 Asia countries · 2000–2025 · Repackaged by Electric Sheep Asia
TL;DR
This dataset contains 203 observations of Number Of Transboundary Animal Breeds Which Have Genetic Resources Secured In Conservation Facilities data across 24 Asia countries, spanning 2000–2025.
About the source
Source: Our… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepasia/asia-owid-number-of-transboundary-animal-breeds-which-have-genetic-resources-secured-in-conserva.MMDecodingTrust-I2T
Overview
This repo contains the image-to-text dataset of MMDT (Multimodal DecodingTrust). This research endeavor is designed to help researchers and practitioners better understand the capabilities, limitations, and potential risks involved in deploying the state-of-the-art Multimodal foundation models (MMFMs). This dataset focuses on the following six primary perspectives of trustworthiness, including safety, hallucination, fairness, privacy, adversarial robustness, and… See the full description on the dataset page: https://huggingface.co/datasets/AI-Secure/MMDecodingTrust-I2T.SECURE-CPSTSECURE-RERTbigvul-securepair