Gotham Device-Level IoT Traffic — 78 Smart City Devices, Federated IDS Benchmark [2026]

Overview

The Gotham Device-Level IoT Network Traffic Dataset is one of the most recent IoT security benchmarks, published to Zenodo in January 2026. Generated using the open-source Gotham testbed, it represents a virtualised smart city environment where 78 heterogeneous IoT devices generate realistic traffic across multiple device classes and communication protocols.

Its defining innovation is the distributed data collection strategy: traffic is captured separately at each IoT device interface rather than aggregated at a central gateway. This preserves the inherently non-IID character of real-world edge data, crucial for validating federated learning-based IDS systems.

The dataset contains 3,134,004 packets and 3,452,465,659 bytes of network traffic across all device captures. Device traffic volumes vary significantly — from low-bandwidth sensors to high-traffic RTSP cameras generating over one million packets.

Column Schema

Key Statistics

• Total Packets: 3,134,004
• Total Bytes: 3,452,465,659
• IoT Devices: 78 heterogeneous smart city IoT devices
• Testbed: Open-source Gotham virtualised smart city environment
• File Format: PCAP and CSV (per-device files)
• Non-IID structure: traffic stored per-device for federated learning
• Published: January 2026 on Zenodo

Use Cases

• Federated learning algorithm development for distributed IoT intrusion detection
• Decentralised IDS evaluation preserving per-device data privacy
• Smart city IoT security analytics across diverse device types and protocols
• Non-IID data benchmarking for privacy-preserving ML in IoT

Source and Attribution

Publicly available on Zenodo (record 14502760), published January 2026. Accompanies research on federated learning-based IDS for heterogeneous IoT environments.