EmuIoT-VT: Emulation-Based Dataset for NIDS in IoT Systems

DescriptionEmuIoT-VT is a network dataset specifically designed for IoT systems. Unlike many existing datasets that rely on synthetic traffic generators, simulations, or traffic replay tools, EmuIoT-VT is constructed by creating virtual replicas of actual IoT devices that closely mirrors real device behavior. The dataset was collected in an fully offilne isolated local network environment, ensuring that the captured traffic is clean.

Key Features 

No Traffic Replays or Generators: All traffic is organically generated through scripted interactions — not replayed or artificially synthesized.

Clean and Labelled Data: Captured in an isolated LAN, ensuring precise ground truth labeling for each flow.

Balanced-by-Design: The dataset includes a uniform distribution of benign and malicious flows to mitigate class imbalance and support robust training.

Supports Binary and Multiclass Classification: Nine total labels including benign and eight distinct attack subtypes across four major attack categories.

Dataset Composition

Total Flows: 28,000 (14,000 benign + 14,000 malicious)

Features: 82 features (features.json: Description of the 82 features extracted from raw PCAP) 

File Format: CSV (benign.csv: records of benign traffic, mal.csv: records of malicious traffic}