Presentation of results in DDoS mitigation at Les Rencontres NetDevOps

Our PI, Gregory, has been invited to present some of the results obtained during Shurok’s Ph.D thesis at “Les Rencontres NetDevOps”, 3rd edition, in Paris. This event is organized by “LeNetDevOps” community, a French domestic initiative led by CNS Communications. It aims at gathering French-speaking people interested in network automation in company environments, including practicioners, vendors and developers, among others.

Gregory’s talk closed a long afternoon of talks and discussions mainly focusing on automation and observability in network deployments, involving speakers from varied sectors including networks and telecommunications, software, services and end-users. Gregory presented the motivation for network automation for threat mitigation, in particular, Distributed Denial of Service (DDoS) attacks, which are continuously increasing in scale. Leveraging the Deep Reinforcement Learning agent developed by Shurok, Gregory demonstrated how a Software-Defined Network can collect statistics on the switches and the traffic they forward to build a state of the network. By exploring different actions on the user traffic, such as redirections or bandwidth manipulations, the agent is able to determine the (optimal) policy to withstand malicious actors that generate DDoS attacks. This approach is able to drastically reduce the effect of the attack, preserving the Quality of Service of user traffic. This research talk, although not completely adhering to this event’s topic, offered a new perspective on autonomic and autonomous networks to a mostly industrial audience.