Datasets:
license: apache-2.0
task_categories:
- text-classification
tags:
- Network Intrusion Detection
- Cybersecurity
- Network Packets
pretty_name: Internet background noise
size_categories:
- 10M<n<100M
configs:
- config_name: default
data_files:
- split: http_2024
path: http-2024-10m.parquet
Internet Background Noise Dataset (Unlabeled Raw Data)
This dataset contains HTTP internet noise data collected by an internet honeypot. It consists of raw, unlabeled network packets, including metadata, payloads, and header information. This data is suitable for training and evaluating machine learning models for network intrusion detection, cybersecurity, and traffic analysis.
HoneyPot repository: hachimi on GitHub.
Dataset Overview
The Internet Background Noise dataset captures raw HTTP traffic from a honeypot designed to simulate typical internet background noise. It includes various network characteristics such as source and destination IP addresses, ports, HTTP headers, payloads, and timestamps. The dataset is provided as raw, unlabeled data to support custom analysis, unsupervised learning, and anomaly detection.
Dataset Schema
The dataset is stored in Parquet format and includes the following fields:
- SrcIp: Source IP address (
BYTE_ARRAY
, UTF-8) - SrcPort: Source port (
INT32
) - DstIpTag: Anonymized destination IP tag (
BYTE_ARRAY
, UTF-8) - DstPort: Destination port (
INT32
) - Country: Country or region name in Chinese (
BYTE_ARRAY
, UTF-8) - IsoCode: Country ISO code (
BYTE_ARRAY
, UTF-8) - Asn: Autonomous System Number (ASN) (
INT32
) - AsnOrg: ASN organization (
BYTE_ARRAY
, UTF-8) - Body: Request body, containing escaped, non-printable characters (may include corrupted binary data) (
BYTE_ARRAY
, UTF-8) - BodyParam: Parsed body parameters in JSON format (
BYTE_ARRAY
, UTF-8) - Error: Boolean flag indicating if an error occurred (
BOOLEAN
) - Method: HTTP method (
BYTE_ARRAY
, UTF-8) - Path: Request path (
BYTE_ARRAY
, UTF-8) - Header: HTTP headers (
BYTE_ARRAY
, UTF-8) (JSON String Map) - TimeStamp: Request timestamp in milliseconds (
INT64
, logical type: TIMESTAMP, unit: MILLIS) - UserAgent: User-Agent string (
BYTE_ARRAY
, UTF-8) - UriParam: Parsed URI parameters in JSON format (
BYTE_ARRAY
, UTF-8) - Data: Complete HTTP request string (escaped binary data) (
BYTE_ARRAY
, UTF-8) - Body_hash: Hash of the request body (SHA-1) (
BYTE_ARRAY
, UTF-8) - Data_hash: Hash of the entire HTTP request data (SHA-1) (
BYTE_ARRAY
, UTF-8) - Path_hash: Hash of the request path (SHA-1) (
BYTE_ARRAY
, UTF-8) - Header_hash: Hash of the request headers (SHA-1) (
BYTE_ARRAY
, UTF-8)
Example Data
Here is a sample of the dataset:
{
"SrcIp": "125.227.119.215",
"SrcPort": 48898,
"DstIpTag": "3j5z5z5z",
"DstPort": 12345,
"Country": "台湾",
"IsoCode": "TW",
"Asn": 3462,
"AsnOrg": "Data Communication Business Group",
"Body": "\\x02\\x00\\x00\\v\\x00\\x00\\t\\x1f\\x01G\\x00\\x12attributes-charset\\x00\\x05utf-8H\\x00...",
"BodyParam": "{}",
"Error": false,
"Method": "POST",
"Path": "/printers/whatever",
"Header": {
"Date": "Fri, 27 Sep 2024 04:17:03 GMT",
"Host": "10.30.6.181:12345",
"Expect": "100-continue",
"User-Agent": "CUPS/2.2.7 (Linux 5.4.0-150-generic; x86_64) IPP/2.0",
"Content-Type": "application/ipp",
"Content-Length": "218"
},
"TimeStamp": "2024-09-27T04:17:04Z",
"UserAgent": "CUPS/2.2.7 (Linux 5.4.0-150-generic; x86_64) IPP/2.0",
"UriParam": "{}",
"Data": "POST /printers/whatever HTTP/1.1\\r\\nContent-Length: 218\\r\\nContent-Type: application/ipp...",
"Body_hash": "685b196d71888a2092bb574dae25d98466209033",
"Data_hash": "13aee0faf1ed4e3cc66c219cab8e415aae7e7c1c",
"Path_hash": "6edc601b36985a576387cd67841d0339a23e3c91",
"Header_hash": "1f0cc56ab844f1cbc137bf8b259c34d63654d553"
}
Notes:
- All destination IP addresses (
DstIpTag
) and Host headers have been anonymized and randomized into the10.0.0.0/8
private IP address space. This ensures that no real-world destination IPs or hosts are included in the dataset. - Source IP addresses (
SrcIp
) remain unchanged for analysis purposes but may include publicly visible IPs collected from internet background noise. - The
Body
andData
fields contain escaped non-printable characters (including newlines and quotes). Some non-printable binary characters may be corrupted due to earlier honeypot design limitations. - The dataset includes both raw binary and textual representations for analysis flexibility.