Abstract

On the cybersecurity market, novel entities – technologies and companies – arise and disappear swiftly. In such a fast-paced context, assessing the survivability of those entities is crucial when it comes to make investment decisions for ensuring the security of critical infrastructures. In this paper, we present a framework for capturing the dynamic relationship between entities of the Swiss cybersecurity landscape. By using open data, we first model our dataset as a bipartite graph in which nodes are represented by technologies and companies involved in cybersecurity. Next, we use patents and job openings data to link the two entities. By extracting time series of such graphs, and by using link-prediction methods, we forecast the (dis)appearance of links. We apply several unsupervised learning similarity-based algorithms, a supervised learning method and finally we select the best model. Our preliminary results show good performance and promising validation of our survivability index. We suggest that our framework is useful for critical infrastructure operators, as a survivability index of entities can be extracted by using the outputs of our models.

Podcast

Scientific Article

Link-Prediction-for-Cybersecurity-Companies-and-Technologies-Towards-a-Survivability-Score

Source : International Conference on Critical Information Infrastructures Security

BibTeX

@inproceedings{anton2021link,
  title={Link Prediction for Cybersecurity Companies and Technologies: Towards a Survivability Score},
  author={Anton Moreno, Santiago and Mezzetti, Anita and Lacube, William},
  booktitle={International Conference on Critical Information Infrastructures Security},
  pages={228--233},
  year={2021},
  organization={Springer}
}