Master thesis; Securing AI-Generated Dependencies

Join our Team

About this opportunity:

Join us for a Master’s thesis exploring slopsquatting, a software supply chain attack in which AI coding assistants recommend non-existent software packages that attackers can register and exploit. You will investigate how this risk affects enterprise software development, particularly in package ecosystems such as npm and PyPI, and examine how secure dependency management and CI/CD controls can support safer AI-assisted development. The work will have direct relevance to Ericsson and other organisations adopting AI tools in software engineering.

What you will do:

You will conduct a literature review and perform controlled experiments with AI coding assistants to measure how frequently they suggest non-existent dependencies. You will assess the exposure of selected language models, analyse how phantom dependencies could enter enterprise development workflows, and evaluate the ability of existing security controls—such as dependency scanners and SBOM-based processes—to detect or prevent them. Based on your findings, you will propose practical mitigation measures, including dependency verification, improved developer guidance and DevSecOps controls. The empirical work will use public or synthetic data rather than confidential Ericsson code, and your findings will be documented in a thesis report and presented to relevant stakeholders.

The skills you bring:

• Background in computer science, cybersecurity, software engineering, artificial intelligence, machine learning or similar


• Good programming skills, preferably in Python or JavaScript, along with a basic understanding of AI, machine learning and large language models


• Knowledge of software supply chain security, secure development practices or CI/CD is beneficial


• Analytical, curious and research-oriented mind, with a proactive and independent approach to problem-solving


• Strong spoken and written English skills, together with the ability to communicate, document findings and collaborate effectively, are essential.