vyykaaa/dataset_all

Web Traffic Normal vs Anomalous

A labeled HTTP request dataset for web traffic classification.

This dataset contains raw HTTP requests collected into two classes:

• anomalous: malicious or attack-like traffic
• normal: benign traffic

Important note about data integrity

The request payloads are kept as raw text. No normalization, decoding, truncation, lowercasing, header removal, or field cleaning was applied when creating the labeled files. The only added field is label.

Files

Recommended files for Hugging Face Datasets:

• web_traffic_labeled_combined.jsonl
• filtered_anomalous_traffic_labeled.jsonl
• normalTrafficTest_clean_labeled.jsonl

Also included:

• web_traffic_labeled_combined.csv
• filtered_anomalous_traffic_labeled.csv
• normalTrafficTest_clean_labeled.csv

Dataset schema

Each example has the following fields:

• text (string): full raw HTTP request, including line breaks
• label (string): one of anomalous or normal

Example

{
  "text": "GET http://localhost:8080/tienda1/index.jsp HTTP/1.1\nUser-Agent: Mozilla/5.0 ...",
  "label": "normal"
}

Class distribution

• anomalous: 2,014
• normal: 36,000
• total: 38,014

Recommended usage

This dataset is suitable for:

• web attack detection
• anomaly detection
• HTTP request classification
• WAF / IDS experimentation
• baseline text classification benchmarks for cybersecurity

Loading with datasets

Load the combined JSONL file

from datasets import load_dataset

dataset = load_dataset("json", data_files="web_traffic_labeled_combined.jsonl")
print(dataset["train"][0])

Load train/test splits from local files

from datasets import load_dataset

dataset = load_dataset(
    "json",
    data_files={
        "train": "web_traffic_labeled_combined.jsonl",
    },
)

Repository layout suggestion

.
├── README.md
├── web_traffic_labeled_combined.jsonl
├── filtered_anomalous_traffic_labeled.jsonl
├── normalTrafficTest_clean_labeled.jsonl
├── web_traffic_labeled_combined.csv
├── filtered_anomalous_traffic_labeled.csv
└── normalTrafficTest_clean_labeled.csv

Limitations

• The labels are binary only: normal and anomalous
• The dataset may contain duplicated request patterns
• The traffic appears to come from a web application testing context and may not cover all modern attack families
• Credentials, hostnames, and request contents should be reviewed before production use

Citation

If you publish work based on this dataset, please cite the dataset repository URL.