Annotated Dataset for Anomaly Detection in a Data Center with IoT Sensors

Area of Science:

• Computer Science
• Network Security
• Data Science

Background:

• Internet of Things (IoT) networks are susceptible to vulnerabilities and failures due to their inherent simplicity.
• Detecting traffic anomalies in IoT networks requires comprehensive, labeled datasets for systematic analysis.
• Existing datasets may not adequately represent real-world IoT network behaviors and potential weaknesses.

Purpose of the Study:

• To introduce the Data Anomaly Detection (DAD) dataset, a novel, labeled resource for IoT network analysis.
• To provide a realistic simulation of IoT network traffic, including specific anomalies.
• To facilitate the development and evaluation of anomaly detection and classification techniques for IoT systems.

Main Methods:

• The DAD dataset was generated using a virtual infrastructure simulating a physical data center with IoT devices.
• NFC smart passive temperature sensors were utilized, connected to a MQTT broker and client nodes.
• Mathematical modeling using time series analysis was employed for data generation and anomaly representation.

Main Results:

• The DAD dataset captures seven days of network activity from a simulated IoT environment.
• It includes three types of anomalies: message duplication, interception, and modification, occurring over five days.
• A detailed feature description is provided for subsequent application of predictive and classification models.

Conclusions:

• The DAD dataset offers a valuable resource for studying IoT network vulnerabilities and anomalies.
• It enables the application of machine learning techniques for anomaly detection and network behavior analysis.
• This dataset supports research aimed at enhancing the security and reliability of IoT infrastructures.