Mathematical Modeling of Computer Network Traffic

školitel: RNDr. Petr Somol, Ph.D.
e-mail: zobrazit e-mail
typ práce: dizertační práce
zaměření: MI_MM, MI_AMSM, II_SIMI
popis: We observe dramatic advances in modeling of image-, audio- and text data as well as of learning frameworks, allowing for solutions to problems only recently deemed too hard to realistically solve - like automated generation of image content descriptions or locating images to place of origin [9], automated play of Go on competitive level [11], learning and transferring artistic style in text or paintings [10]. Large part of theses successes is due to improvements of models allowing for finding structural information in data and reliably estimating the respective distributions. So far, not many such techniques have been developed in context of computer network traffic. Many techniques [1, 2, 3, 4, 5] give promise of tackling the problem of network data modeling, though their current state is insufficient to enable reliable application in computer network domain. Developing these techniques and finding new ones in the area of computer networks has great potential to enable significant advances in computer security [6, 7], computer network administration [8], and eventually making Internet of Things a reality. As part of work on this thesis the student is expected to devise models on top of data representing network traffic of very large real networks. The primary aim is to enable discovery of anomalies and their classification in order to reveal possible malicious activity of either malicious software or human adversaries. Computer network data has difficult properties: imprecise definition of classes, difficult dimensionality-to-sample size ratio, imbalance, missing feature values, difficult structure, unclear local context. Moreover, the data can be almost intractably large. The key to success in modeling the difficult computer network data is in finding formally correct generalising models of proven properties as well as procedures for their parameters' optimization.
literatura: [1] Amores, Jaume. \"Multiple instance classification: Review, taxonomy and comparative study.\" Artificial Intelligence 201 (2013): 81-105.
[2] Bengio, Yoshua, Aaron Courville, and Pierre Vincent. \"Representation learning: A review and new perspectives.\" Pattern Analysis and Machine Intelligence, IEEE Transactions on 35.8 (2013): 1798-1828.
[3] Pimentel, M. A., Clifton, D. A., Clifton, L., & Tarassenko, L. (2014). A review of novelty detection. Signal Processing, 99, 215-249.
[4] Jain, A. K., Duin, R. P., & Mao, J. (2000). Statistical pattern recognition: A review. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 22(1), 4-37.
[5] Frénay, B., & Verleysen, M. (2014). Classification in the presence of label noise: a survey. Neural Networks and Learning Systems, IEEE Transactions on, 25(5), 845-869.
[6] Grill, Martin, Tomáš Pevný, and Martin Rehak. \"Reducing false positives of network anomaly detection by local adaptive multivariate smoothing.\" Journal of Computer and System Sciences (2016).
[7] Bartos, Karel, and Michal Sofka. \"Robust Representation for Domain Adaptation in Network Security.\" Machine Learning and Knowledge Discovery in Databases. Springer International Publishing, 2015. 116-132.
[8] Rehak, M., Pechoucek, M., Grill, M., Stiborek, J., Bartoš, K., & Celeda, P. (2009). Adaptive multiagent system for network traffic monitoring. IEEE Intelligent Systems, (3), 16-25.
[10] Gatys, Leon A., Alexander S. Ecker, and Matthias Bethge. \"A neural algorithm of artistic style.\" arXiv preprint arXiv:1508.06576 (2015).
naposledy změněno: 27.05.2016 08:35:01

za obsah této stránky zodpovídá: Ľubomíra Dvořáková | naposledy změněno: 12.9.2011
Trojanova 13, 120 00 Praha 2, tel. 224 358 540, pevná linka 224 923 098, fax 234 358 643
České vysoké učení technické v Praze | Fakulta jaderná a fyzikálně inženýrská | Katedra matematiky