Samenvatting
We discuss the detection of botnets by means of hypothesis testing in a random graph model. For this, we construct a test that detects the presence of a botnet in a complex network and prove that this test has an asymptotic risk of zero when the size of the botnet is reasonably large.The random graph model we introduce differentiates between bots and real vertices by means of underlying geometry and more specifically assume that bots will form connections with other vertices according to the Chung-Lu model, while real vertices form connections with other real vertices as in geometric inhomogeneous random graphs. Based on this model, we construct the based on the maximal number of independent neighbours a vertex has, i.e. the isolated star number. In specific we prove in this report that bots tend to have larger Isolated Star Numbers than real vertices.
| Datum prijs | 11 mrt. 2021 |
|---|---|
| Originele taal | Engels |
| Begeleider | Julia Komjathy (Afstudeerdocent 1) & Joost Jorritsma (Afstudeerdocent 2) |