[PhD position 2021] Building Self-Driven Network Functions

Keywords : Artificial Intelligence, 5G Networks, Machine Learning, Network Intents

 Contact:

Abdelkader LAHMADI (lahmadi@loria.fr),

Michael RUSINOWITCH (michael.rusinowitch@inria.fr)

Scientific Context:

As computer networks becomes more complex over time, the management of their functions including filtering, routing, load balancing, traffic engineering and fire-walling become an increasingly important concern for the organization operating it. More recently, these networks are becoming software defined (SDN) and virtualized (NFV) to make their deployment and management easier. However they still rely on network administrators and human defined policies to configure them and achieve the required operations. They are still not able to make the decisions when facing a new threat, a link failure or sudden peaks of network traffic.

Recent trends in networking are to rely on Machine Learning (ML) techniques for the control and the operations of these functions.  Most of the existing works are either using ML for building network analytics [1,2] or optimizing routing functions [3]. Other works are using ML techniques for inferring network behavior, in particular by observing forwarding behaviors [4,5].

In this context, we propose a PhD subject to improve the learning capabilities of network functions [6] and make them more autonomous with advanced decision making, closing the loop between the network behavior and the required intents to meet the desired operations.

Subject and Goals:

In this PhD, we focus on building native ML networking functions that rely on learning the required heuristics to drive themselves and meet their operations.

A first objective is to design an automated learning architecture for distributing  filtering functions over the network switches. Filtering functions are critical components for enforcing security policies e.g., by blocking or redirecting malicious traffic. As switches’ resources are often strongly limited, filtering functions should be optimally positioned in the network, avoiding redundancy and traffic slowdown. This problem is NP-hard but several approximation algorithms and heuristics to solve it have been proposed in the literature. However, given a network configuration it is not obvious to understand  which one is the best suited. For a flexible, fully automated solution we propose to proceed from reinforcement learning methods and graph embeddings.

The above objective can be addressed only if policies have been formally specified by administrators. However it is often the case in practice that these specifications are not available [5]. Therefore a second objective is to mine the required filtering policies from the network behavior and close the loop with the reinforcement learning mechanism.

A third objective is the development of a platform for experimenting different ML architectures and network functions placement problems. This development will be carried out by leveraging Mininet emulator and P4 network programming language.

References:

[1] Changhoon Kim, Anirudh Sivaraman, Naga Katta, Antonin Bas, Advait Dixit, and Lawrence J Wobker. In-band network telemetry via programmable dataplanes. In ACM SIGCOMM, 2015.

[2] Alexander Clemm, Mouli Chandramouli, and Sailesh Krishnamurthy. Dna: An sdn framework for distributed network analytics. In Integrated Network Management (IM), 2015 IFIP/IEEE International Symposium on, pages 9–17. IEEE, 2015.

[3] Shih-Chun Lin, Ian F Akyildiz, Pu Wang, and Min Luo. QoS-Aware Adaptive Routing in Multi-layer Hierarchical Software Defined Networks: A Reinforcement Learning Approach. In 2016 IEEE International Conference on Services Computing (SCC), pages 25–33. IEEE, June 2016.

[4] Rüdiger Birkner, Dana Drachsler-Cohen, Laurent Vanbever, and Martin Vechev. 2018. Net2Text: query-guided summarization of network forwarding behaviors. In Proceedings of the 15th USENIX Conference on Networked Systems Design and Implementation (NSDI’18). USENIX Association, USA, 609–623.

[5] Ali Kheradmand. 2020. Automatic Inference of High-Level Network Intents by Mining Forwarding Patterns. In Proceedings of the Symposium on SDN Research (SOSR ’20). Association for Computing Machinery, New York, NY, USA, 27–33. DOI:https://doi.org/10.1145/3373360.3380831

[6] Nick Feamster and Jennifer Rexford. 2018. Why (and How) Networks Should Run Themselves. In Proceedings of the Applied Networking Research Workshop (ANRW ’18). Association for Computing Machinery, New York, NY, USA, 20.

Hosting teams:

The PhD position is proposed by the RESIST (https://team.inria.fr/resist/) and PESTO (https://team.inria.fr/pesto/) teams of LORIA and Inria Nancy-Grand Est research lab.

Skills and profile:

  • Required qualification: Master in Computer Science
  • Knowledge and skills in the following fields will be appreciated: networking, security, machine learning, big data, programming (C, python)

Additional information:

Duration: 3 years

Starting date: between Oct. 1st 2021 and Jan. 1st 2024
How to apply:

Send the following documents to lahmadi@loria.fr and michael.rusinowitch@inria.fr

– CV;

– a motivation letter;

– your degree certificates and transcripts for Bachelor and Master (or the last 5 years if not applicable).

– Master Thesis (or equivalent) if it is already completed, or a description of the work in progress, otherwise;

– all your publications, if any (it is not expected that you have any).

– At least one recommendation letter from the person who supervises(d) your Master thesis (or research project or internship); you can also send at most two other recommendation letters. The recommendation letter(s) should be sent directly by their author to the prospective PhD advisor.

Logo d'Inria