Another opportunity of dissemination for our Ph.D student, Solayman, who is bound to present his last accepted paper at the 19th International Conference on Risks and Security of Internet and Systems (CRiSIS 2024), which be held in Aix-en-Provence from November 26th. The paper, entitled FREIDA: A Concrete Tool for Reproducible Evaluation of IDS using a Data-driven Approach, presents the implementation of our FREIDA framework, which enables comprehensive assessments of ML-based IDS. The paper further exemplifies its applicability by evaluating IDS according to two aspects: their detection effectiveness, and robustness to adversarial attacks. Solayman will be presenting on the starting day of CRiSIS (see programme), so come and discuss!

Our two Ph.D students, Shurok and Solayman, are invited to present the results of their work in the GRIFIN project during the kickoff of TURFU-NET. The kickoff meeting will take place on November 22nd in Illkirch-Graffenstaden (see program). Shurok will present her reinforcement learning approach to mitigate DDoS attacks in a talk entitled DDoS Mitigation while Preserving QoS: A Deep Reinforcement Learning-Based Approach while Solayman will discuss the FREIDA framewok in a talk entitled Towards Reproducible Evaluations of ML-Based IDS Using Data-Driven Approaches. TURFU-NET (Transcending the Usual Rationale for the Future of Ubiquitous NETworks) is an ANR TSIA project, a call specific to Artificial Intelligence (AI) topics. Coordinated by Quentin Bramas, from the University of Strasbourg, it aims at redefining network management and optimization through the integration of cutting-edge neurosymbolic AI. The project partners come from research teams located at Strasbourg (iCube laboratory), Mulhouse (IRIMAS laboratory, Université de Haute-Alsace) and Paris (LIP6, Sorbonne University).

The second installment of ARTMAN, the workshop on Recent Advances in Resilient and Trustworthy MAchine learniNg is progressing towards completion. As a reminder, it will be held before ACSAC, in Hawai on December 9th, 2024. The program is now available online on both ACSAC’s and ARTMAN’s websites. The program features 4 technical sessions showcasing 12 selected papers on a variety of topics dealing with ML for cybersecurity, robustness, privacy and safety for ML systems, and attacks to ML algorithms. This year’s ARTMAN is highlighted by two outstanding keynotes: Melek Önen, professor at Eurecom, will be speaking in the morning about Customized attacks and defense strategies for robust and privacy-preserving federated learning; and Christian Wressnegger, assistant professor at Karlsruhe Institute of Technology, will intervene during the afternoon on the topic of Prospects and Limits of Explainable AI in Computer Security. Do not wait to register at ACSAC (do here, early ends after November 15th) and join us for this exciting workshop.

Solayman, our PhD candidate, will be presenting a demo on the FREIDA framework, and its current implementation during the poster/demo session at ACM CCS, happening right now in Salt Lake City, USA. The demo will feature a walk-through the different stages of the ML-IDS evaluation pipeline with a focus on the reproducibility on one hand, and data manipulation on the other hand. As demo booths are not fixed, please look out for Solayman’s during Poster and Reception on Wednesday 16th at 4:30pm in Salons G, H and I. The demo is entitled Towards Reproducible Evaluations of ML-Based IDS Using Data-Driven Approaches.

Solayman, our PhD candidate, will be presenting his work on the FREIDA framework, an ongoing effort on assisting researchers with a data-driven evaluation of ML-based intrusion detection systems. He is invited to talk in front of the SuperviZ project members for their plenary meeting in Nancy on September 27th. His work is of interest to SuperviZ’ working group 5 on IDS evaluation. The work will later be discussed among members of this working group. This work is also bound to be presented later next month at an international conference, which will be disclosed in a later post.

After a successful first installment, co-located with the 30th ACM CCS last year in Copenhagen, ARTMAN is back in 2024. With a somehow broader scope, the rebranded workshop on Recent Advances in Resilient and Trustworthy MAchine learniNg is now joining forces with the 40th edition of ACSAC, held in Hawai in December 2024. The Call for Papers is out and available on ARTMAN’s website. The submission deadline has been extended to September 15th, 23:50 AoE. Accepted papers are likely to be published to an international journal, as co-chairs are actively considering a suitable venue.

Our PI, Gregory, was invited to open the CNRS’ GDR SI Summer School held in Bourges (France), Cyber in Berry 2.0, by a talk on his experiences in ML-based network intrusion detection. The keynote, entitled Learning-based Network Intrusion Detection: Are We There Yet?, covers topics at the intersection of the Machine Learning and Intrusion Detection domains, summing up both state of the art and Gregory’s own research, including latest endeavours in NIDS evaluation, as carried out in GRIFIN. Despite a postponed talk late in the afternoon, Gregory received positive feedback and offered the attendees a nice aperitive, discussing issues and perspectives in the ML-based NIDS research scene. Hoping to have inspired students to follow in his trail…

Today, after a quick trip to St. Louis, our PI, Gregory, will be delivering a presentation on DDoS Mitigation while Preserving QoS: A Deep Reinforcement Learning-Based Approach. This presentation prepared by Shurok for her accepted submission at SecSoft ‘24 will be presented on her behalf, as she could not attend herself. It takes place as NetSoft ‘24 is coming to an end with this last day workshop on Cyber-Security in Software-defined and Virtualized Infrastructures (program).

Shurok will be giving a talk on her upcoming presentation at SecSoft ‘24. She will be discussing the contents of her accepted research paper during next CoaP seminar on June 17th from 10am in room 3.A213. This will be the opportunity to get a first feedback from other research colleagues around our campus.

Tomorrow marks the start of another edition of the European Interdisciplinary Cybersecurity Conference (EICC). There, one output of the GRIFIN project will be showcased as collaboration with Japanese scholars from The University of Tokyo and Toyo University. The joint work conducted by Satoshi Okada, a Ph.D candidate who stayed for 6 months with us in 2023, introduces a methodology and related experiments with leveraging explainable AI metrics such as Integrated Gradients to drive the process of looking for adversarial examples in the problem space that are able to circumvent a network intrusion detector. In this work, Satoshi demonstrates successful yet realistic bypasses of an NIDS trained the CIC-IDS2017 dataset for a couple of attack classes, with little effort. He then validates that the generated adversarial network packets work in practice by attacking the detector in a virtual testbed. Please attend his talk during Session 1B on Machine Learning and Security in Room A from 11:15 to 12:35 on the first day (June 5th).