| --- |
| license: apache-2.0 |
| task_categories: |
| - tabular-classification |
| tags: |
| - cybersecurity |
| - intrusion-detection |
| - server-security |
| - honeypot |
| - log-analysis |
| pretty_name: DXA-Guard Intrusion Detection Benchmark Dataset |
| size_categories: |
| - 1K<n<10K |
| --- |
| |
| # DXA-Guard Intrusion Detection Benchmark Dataset |
|
|
| Official benchmark dataset for **[DXA-Guard](https://github.com/deployxa/dxa-guard)** and the **[deployxa/dxa-guard](https://huggingface.co/deployxa/dxa-guard)** model, published by **[Deployxa](https://huggingface.co/deployxa)**. |
|
|
| ## Overview |
| This dataset contains real-world and synthetic honeypot security events captured across Linux VPS instances. Events include SSH authentication attempts (brute force, dictionary, credential stuffing) and web server requests (vulnerability scanners, path traversal, SQL injection, webshell uploads, and benign traffic). |
|
|
| ## Splits |
| - `train.jsonl`: Training event stream partitioned chronologically. |
| - `test.jsonl`: Independent held-out test evaluation set with distinct IP addresses and attack patterns. |
|
|
| ## Schema |
| Each line is a JSON object with: |
| - `timestamp`: RFC3339 timestamp string |
| - `source_ip`: Source IP address |
| - `action`: Event type (`ActionSSHLogin`, `ActionHTTPRequest`, etc.) |
| - `user`: SSH username attempted (if SSH) |
| - `method`: HTTP method (`GET`, `POST`, etc.) |
| - `path`: HTTP request path |
| - `status_code`: HTTP response status code |
| - `label`: `"attack"` or `"benign"` |
| - `rule_match`: Matching detection rule identifier (if attack) |
|
|
| ## Associated Model |
| Model weights: [`deployxa/dxa-guard`](https://huggingface.co/deployxa/dxa-guard) |
|
|
