Malicious Website Detection with AI
Detecting phishing and malicious websites is a critical task. This topic explores how AI systems can analyze web content, metadata, and context.
Machine Learning/AI and Security
Open Topics
The following contains currently open thesis topics in our areas of research. If you are interested in a topic or have a thesis idea of your own that might fit our research interests, please contact us directly.
Membership inference attack
Analyse and evaluate attack scenarios for membership inference, analyse existing attack and defense patterns, and develop new mechanisms.
Machine Learning/AI and Security
Mobile app analysis
- Extracting cloud storage URLs from mobile apps
- Extracting cloud authentication from mobile apps and testing for authorization issues (Cognito Identity Provider)
- Using LLMs to classify the purpose of app secrets
- Extracting and analyzing JWT and private keys
Please contact Sebastian Schrittwieser.
Software Protection
Mobile App Analysis Automation
- Building an end-to-end pipeline: APK/IPA acquisition, static extraction, and dynamic instrumentation
- Automating Frida/objection scripting for common analysis tasks across heterogeneous apps
- Scaling secret extraction and classification (cloud URLs, JWTs, keys) to thousands of apps
- Continuous monitoring:…
Software Protection
Multi-Agent Collaboration, Memory, and Coordination in Cybersecurity
Complex cybersecurity tasks require collaboration between multiple agents. This topic explores how agent-based systems can be coordinated and supported through shared memory and knowledge structures.
Machine Learning/AI and Security
Obfuscation detection
- Obfuscation detection for JavaScript
- Obfuscation detection for cross-platform frameworks (e.g., Flutter)
- Obfuscation detection in Android apps
- Obfuscation detection in binary code
- Scanning for obfuscation usage in GitHub, e.g., Tigress or Proguard
Please contact Sebastian Schrittwieser.
Machine Learning/AI and Security
Other attacks, e.g. Model stealing, Model inversion
- Model stealing: “Stealing Machine Learning Models via Prediction APIs” & video
- Model inversion: “Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures”
Please contact Rudolf Mayer.
Machine Learning/AI and Security
Physically Unclonable Functions in commodity hardware
- Surveying PUF candidates in commodity components (e.g., GPU components)
- Measuring entropy, reliability, and reproducibility of PUF responses across temperature and aging
- Evaluating resistance to cloning and machine-learning modeling attacks
Please contact Sebastian Schrittwieser.
Software Protection
Privacy enhancing technologies
- Anonymous credentials and their deployment trade-offs
- Zero-knowledge proof systems in applied setting
- Metadata-less messaging (e.g, Katzenpost)
- Tor and mix networks
Please contact Aljosha Judmayer.
Security and Privacy of Decentralized Systems and DLTs