Title: SPARTAN: A Persona-Driven Smart-Home IoT-Security Dataset for APT Detection

Duration: 1 year

Research Area: IoT Security, Machine Learning

SPARTAN takes a different approach to IoT security research by focusing on realistic user behavior.
Using a “Persona” method, the project simulates diverse household activities and daily routines to
generate authentic benign traffic via the Rohde & Schwarz TTDAS system. This realistic baseline is
combined with APT attack phases to train robust Machine Learning models. An important aspect is
knowledge transfer: a Summer 2026 master’s seminar will teach students dataset creation. Students
will develop their own projects and at least one master’s thesis, actively contributing to the ongoing
research and bridging academic training with industrial application.

Aims

The project aims to create a realistic, structured, and parametrizable dataset that combines diverse
household behavior with comprehensive APT attack scenarios. The goal is to enable effective training
and evaluation of machine learning-based APT detection methods for smart-home environments.

Problems

Existing IoT security datasets fail to capture realistic usage scenarios of smart home devices from real
test-beds and lack realistic benign traffic from multiple users. But complex APT activities often blend
into legitimate behavior, making detection difficult.

Technology

The project employs the Test and Training Data Automation System (TTDAS) by Rohde & Schwarz for
automated device interaction control on Android and iOS platforms. The testbed includes diverse
smart-home devices. Network traffic is captured using the ProfiShark 1G.

Outlook

SPARTAN advances smart-home security by providing quality training data for detection models. The
flexible, persona-based approach supports testing under different household setups and attack
scenarios. Future research directions include detecting privacy attacks, integrating device logs for
better context, and expanding to additional attack types and IoT protocols. The student projects from
the seminar may also contribute new research directions.

Publications

• V. Jüttner, M. Vargas, C. Jahn, M. Mieth, E. Buchmann “SPARTAN: A Persona-Driven Smart-
  Home IoT-Security Dataset for APT Detection” (submitted)

Team

Lead

Prof. Dr. Erik Buchmann

Team Members

• Victor Jüttner
• Arne Roszeitis

Partners

• ipoque GmbH – A Rohde&Schwarz Company