Hakan Kayan

Hakan Kayan

PhD Student at Cardiff University / Prifysgol Caerdydd

Welcome

About Me

I am a researcher specializing in IoT-based anomaly detection, edge AI, adversarial robustness, and embedded ML. My current focus is on designing lightweight anomaly detection pipelines for industrial cyber-physical systems, particularly in real-time inference on embedded devices such as Nicla Sense ME.

My research has contributed to open-source model compression, real-time AI-powered cybersecurity solutions, and robust ML deployments in constrained environments. I actively work with TinyML, ONNX, PyTorch, TensorFlow Lite Micro, and edge-to-cloud IoT pipelines.

I enjoy working with Linux, microcontrollers, and edge/fog computing to develop scalable ML-powered embedded solutions.

📩 Contact me: kayanh@cardiff.ac.uk

Research Interests

  • Edge & Embedded Machine Learning – TinyML, STM32, Nicla Sense ME.
  • Adversarial Robustness & Model Compression – Quantization, pruning, distillation.
  • Anomaly Detection for Cyber-Physical Systems – Industrial robotic arms, IoT.
  • Neural Networks & AI Optimization – 1D-CNN, LSTM, XGBoost, One-class SVM.
  • Industrial IoT & Edge AI – BLE, MQTT, RTDE, real-time AI monitoring.
  • Sensor Data Processing – IMU fusion, Madgwick/Mahony filters, feature extraction.
  • ML Pipelines & Automation – TensorFlow Lite Micro, ONNX, PyTorch.

Education

  • Ph.D. in Computer Science & Informatics (Expected 2025)
    Cardiff University
    • Developed an open-source anomaly detection pipeline for embedded systems.
    • Adapted TensorFlow Lite Micro for Nicla Sense ME and optimized firmware to free 20% RAM.
    • Achieved 98%+ accuracy in real-time robotic arm anomaly detection using LSTM & 1D-CNN.
    • Integrated BLE-based IoT monitoring with Grafana, InfluxDB, and Google Cloud.
  • M.S. in Information Security (Distinction, 2019)
    University of Surrey
    • Developed a network-based intrusion detection system (IDS) on Raspberry Pi by porting Suricata.
  • B.S. in Electrical & Electronics Engineering (1st in Department, 2017)
    Izmir University of Economics
    • GPA: 3.93/4.0
    • Specialized in real-time signal processing & embedded AI.

Work Experience

  • Senior Research Associate (Oct 2024 - Present)
    Cardiff University
    • Optimizing MobileNetV3 robustness against adversarial attacks using Brevitas & Torch-Pruning.
    • Developing efficient model compression techniques for edge-based industrial AI applications.
    • Investigating the relationship of adversarial robustness and model compression.
  • Research Associate (Apr 2024 - Oct 2024)
    Cardiff University
    • Developed a C++ ML port for an STM32-based PCB with a BioZ IC for body hydration measurement.
  • Chief Scientific Officer (CSO) (Apr 2022 - Mar 2023)
    UKRI – CyberASAP
    • Developed AI/ML algorithms for CASPER, an IoT-based anomaly detection system.
    • Built a cross-platform mobile app using Flutter for real-time anomaly alerts.
  • Research Assistant (Sep 2021 - Sep 2022)
    Cardiff University
    • Developed hands-on IoT labs for MSc and BSc students.
    • Integrated microcontroller programming with edge-to-cloud AI/ML systems.
  • Project Engineer (Jun 2015 - Sep 2017)
    Embryonix Technology Transfer Office
    • Supervised 50+ AI/ML projects, guiding teams in developing edge AI-based solutions.

Skills

  • Programming: Python, C++, ONNX, TensorFlow, PyTorch
  • Embedded Systems: TinyML, STM32, Arduino, ESP32, Raspberry Pi
  • IoT & Edge AI: BLE, MQTT, RTDE (Universal Robots)
  • Machine Learning: Model compression, anomaly detection, adversarial robustness
  • Cloud & DevOps: Docker, Grafana, InfluxDB, Google Cloud

Languages

  • Turkish (Native)
  • English (Professional Fluency)
  • Japanese (N4 - Pre-Intermediate)

Selected Publications

  • Kayan, H., Heartfield, R., Rana, O., Burnap, P., Perera, C., 2024.
    CASPER: Context-Aware IoT Anomaly Detection System for Industrial Robotic Arms.
    ACM Transactions on Internet of Things. DOI

  • Kayan, H., Majib, Y., Alsafery, W., Barhamgi, M., Perera, C., 2021.
    AnoML-IoT: An End-to-End Reconfigurable Multi-Protocol Anomaly Detection Pipeline for IoT.
    Internet of Things, 16, p.100437.

  • Kayan, H., Nunes, M., Rana, O., Burnap, P., Perera, C., 2021.
    Cybersecurity of Industrial Cyber-Physical Systems: A Review.
    ACM Computing Surveys (CSUR).

For a full list, check my Google Scholar.

PhD Research & Open-Source Contributions

A significant portion of my PhD research, including real-time anomaly detection in industrial robotic arms using TinyML, is available in the following repository:

➡️ Real-time Anomaly Detection in Industrial Robotic Arms via TinyML

This repository contains:

  • TinyML-based anomaly detection models (1D-CNN, LSTM, XGBoost, One-class SVM)
  • Edge AI deployment on Nicla Sense ME
  • RTDE integration with Universal Robots
  • BLE-based real-time streaming and monitoring
  • Data preprocessing, feature extraction, and model compression techniques

For more details on my PhD work, check my Google Scholar.