Projects

The goal of SecMAS is to develop and validate secure communication and control mechanisms for multi-agent AI systems to prevent unauthorized access, privilege escalation, and the spread of harmful inputs between agents. The project aims to establish a scientific foundation for the safe deployment of agentic AI in high-risk environments such as cybersecurity and automated penetration testing.

AI-generated content and online misinformation are making it harder to know what is real and trustworthy on the internet. The InReality platform addresses this problem by recording digital content as soon as it is created by verified and trusted sources, and by making sure the content cannot be secretly changed later. Even if parts of the system are hacked or manipulated, any tampering can still be detected.

In this project, experts from the Alexandra Institute have joined forces with the Danish startup PrivacyMate to close the security gaps introduced by AI agents. The outcome will be an AI agent capable of handling data securely and accurately, as well as a tool that can be deployed internationally.

Data normally flows safely through networks, but cyberattacks can subtly disrupt it—often unnoticed. Traditional NIDS based on synthetic data may miss such threats. This project combines machine learning with logic and differentiable logics trained on real data to create monitoring systems with digital intuition that detects even well-hidden attacks.

The project focuses on improving fraud detection in digital payments. Traditional systems typically analyze transactions one by one, which creates blind spots that organized fraud groups can exploit by spreading small activities across many accounts. It aims to use more advanced methods that detect patterns across transactions and relationships between users. The goal is to make fraud detection more accurate, faster, and better at identifying coordinated fraud attacks in real time.

This project aims to develop an analysis tool that enables developers to detect and mitigate vulnerabilities in modern hybrid apps. The project will explore how static code analysis can be combined with dynamic analysis to detect sensitive data flows, for example from Java to JavaScript.

In the quantum era, computers will solve mathematical problems far more efficiently, making post-quantum cryptography (PQC) essential. The challenge is to make PQC as practical and seamless as today’s systems. The project addresses that challenge by adapting threshold-based key management—where keys are shared across multiple parties—to remain secure in a quantum-resistant future.

Satellites are increasingly vulnerable to cyberattacks, especially as they rely on interconnected ground systems and are often developed by smaller companies with limited cybersecurity resources. This project addresses this new reality, where satellites sit on the digital frontline, and sophisticated attacks must be quickly detected and mitigated.

The project will develop an intelligent monitoring system for OT networks, which are often classified as critical infrastructure. The system continuously checks equipment security, detects suspicious behavior, and alerts teams when intervention is needed, giving companies and authorities an early, clear overview of their security posture.

This project aims to provide the first complete security analysis of MLS at both the cryptographic and software implementation levels, similar to what already exists for TLS and Signal.

The project delivers a comprehensive framework that assesses risks and ensures traceability throughout the entire 3D printing process. It also incorporates encryption allowing partners to share designs without revealing intellectual property. The goal is to enable operators and suppliers to document high-quality production and detect deviations before they escalate into costly errors.