Inference and parameter estimation in bayesian change point models
| dc.contributor | Graduate Program in Computer Engineering. | |
| dc.contributor.advisor | Cemgil, Ali Taylan. | |
| dc.contributor.author | Yıldız, Çağatay. | |
| dc.date.accessioned | 2023-03-16T10:02:42Z | |
| dc.date.available | 2023-03-16T10:02:42Z | |
| dc.date.issued | 2017. | |
| dc.description.abstract | In this work, we present a Bayesian change point model that identifies the time points at which a time series undergoes abrupt changes. Our model is a hierarchical hidden Markov model that treats the change points and the dynamics of the data stream as latent variables. We describe a generic generative model, forward-backward recursions for exact inference and an expectation-maximization algorithm for hyper parameter learning. The model specifications discussed here can sense the changes in the state of the observed system as well as in the intensity and/or the ratio of the features. In addition to investigating the change point algorithm in generic notation, we also give an in-depth analysis and appropriate implementation of a particular model specification, namely, Dirichlet-Multinomial model. We present a novel application of the model: Distributed Denial of Service (DDoS) attack detection in Session Initiation Protocol (SIP) networks. In order to generate DDoS attack data, we build a network monitoring unit and a probabilistic SIP network simulation tool that initiates real-time SIP calls between a number of agents. Using a set of features extracted from target computer’s network connection and resource usage statistics, we show that our model is able to detect a variety of DDoS attacks in real time with high accuracy and low false-positive rates. | |
| dc.format.extent | 30 cm. | |
| dc.format.pages | xiii, 61 leaves ; | |
| dc.identifier.other | CMPE 2017 Y56 | |
| dc.identifier.uri | https://digitalarchive.library.bogazici.edu.tr/handle/123456789/12338 | |
| dc.publisher | Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2017. | |
| dc.subject.lcsh | Bayesian field theory. | |
| dc.title | Inference and parameter estimation in bayesian change point models |
Files
Original bundle
1 - 1 of 1