Data mining for discovering anomalies and aiding analysis of telecommunication networks

Contact

René Quiniou (rene.quiniou@inria.fr), Laurence Rozé

Keywords

data mining, time series, telecommunication network data, anomaly discovery

Context

Applications running through telecommunication networks, e.g. those related to e-commerce, bank services or business, in general, require a high quality of serviice (QoS) to avoid unexpected bad financial consequences, both for companies and their customers. Detecting or predicting troubles early is therefore of high importance to provide robust network services with high QoS. With the advent of cloud computing and the dissemination of computing devices (computers, smartphones, etc.) the nework traffic keeps on increasing  which in turn increases the difficulty of detection and prediction tasks.

Problem

Data mining gathers diverse computational methods such as clustering or pattern mining. Such methods and related algorithms are supposed to scale up for processing big datasets. Data mining has been used for detecting intrusion or  predicting QoS degradation. However, the volume and the complexity of data have increased so much that original data mining methods should be revised  to face the challenges related to “Big Data”.

Task

In the case of network monitoring, data arrive as time series in continuous streams. The goal of the internship is, first, to make a survey of  data mining methods proposed to process such data. Second, the student will focus on the kowledge discovery task from big sets of times series; The data will be provided by a company specialized in network data analysis, with which Dream is beginning to collaborate. The student will select the most promising algorithms for processing voluminous datasets. Then, he will adapt some of these algorithms, or design new ones, to cope with voluminous data for network data analysis and anomaly discovery and detection.

Applicant

The applicant should have a good background in data mining, especially in pattern-mining. Some knowledge about telecommunication networks will be a bonus.

References

Manish Joshi and Theyazn Hassn Hadi. A Review of Network Traffic Analysis and Prediction Techniques. 2015 (http://arxiv.org/abs/1507.05722)
Marie-Odile Cordier, Roberto Micalizio, Sophie Robin, Laurence Rozé. Adapting Web Services to Maintain QoS Even When Faults Occur. ICWS 2013: 403-410
Wei Wang, Thomas Guyet, René Quiniou, Marie-Odile Cordier, Florent Masseglia, Xiangliang Zhang. Autonomic intrusion detection: Adaptively detecting anomalies over unlabeled audit data streams in computer networks. Knowl.-Based Syst. 70: 103-117 (2014)
Thomas Guyet, René Quiniou. Extracting Temporal Patterns from Interval-Based Sequences. IJCAI 2011: 1306-1311 (2011)