Bachelor and Master Theses

Also for M.Sc in Cybersecurity 

Privacy-preserving AI is essential for domains like healthcare, finance, and identity management. Traditional machine learning requires centralized data access, which conflicts with GDPR and other regulations. Modern techniques such as Federated Learning (FL), Differential Privacy (DP), and Secure Multiparty Computation (SMPC) allow training without exposing raw data.
However, these approaches introduce trade-offs between accuracy, privacy guarantees, and computational overhead. There is a lack of comparative studies evaluating their performance on realistic datasets.

Problem Statement

How can we design and evaluate privacy-preserving AI pipelines that maintain strong privacy guarantees while minimizing performance degradation?

ML/DL Techniques

• Random Forest / Gradient Boosting
• Autoencoders for anomaly detection (zero-day threats)
• 1D CNN / LSTM / Transformer-based IDS models

Tools

• Python, PyTorch/TF
• Wireshark/Tshark for traffic analysis
• Splunk/ELK for SOC simulation (optional)