On March 25 and 26 , the University of Luxembourg is joining forces with the Cyberexcellence consortium to assess the possibility of jointly undertaking ambitious European projects. More than 40 members of the consortium will be attending the workshop, which is an opportunity for researchers to discuss their research productively, as well as a great teambuilding experience.

Speakers/ abstracts :

Jules Dejaeghere (UNamur, WP4) - Rethinking updates in anonymous communication networks

Anonymous communication networks enable people to browse anonymously the Internet. Tor, one popular anonymous communication network that has been around for more than two decades, has been facing challenges when it comes to updating nodes in the network. With new attacks being discovered against Tor, the developers come with new versions of the software to preserve the online privacy of Tor's users. However, shipping and applying new versions to a network of ~7000 nodes run by volunteers across the planet remains challenging.

This presentation walks through a new software paradigm for handling software updates as part of the usual software execution. Using a portable compilation target and just-in-time compilation, this paradigm can ease the deployment of updates for software projects, particularly for distributed software with a centralized development team like Tor.

Tom Rousseaux (UClouvain, WP1) - Network Simulator-centric Compositional Testing

This presentation introduces a novel methodology, Network Simulator-centric Compositional Testing (NSCT), to enhance the verification of network protocols with a particular focus on time-varying network properties. NSCT follows a Model-Based Testing (MBT) approach. These approaches usually struggle to test and represent time-varying network properties. NSCT also aims to achieve more accurate and reproducible protocol testing. It is implemented using the Ivy tool and the Shadow network simulator. This enables online debugging of real protocol implementations. A case study on the picoquic implementation of QUIC is presented, revealing an error in its compliance with a time-varying specification. This error has subsequently been rectified, highlighting NSCT’s effectiveness in uncovering and addressing real-world protocol implementation issues. The presentation underscores NSCT’s potential in advancing protocol testing methodologies, offering a notable contribution to the field of network protocol verification.

Maxime Goffard (ULiege, WP2) - MSTG: A Flexible and Scalable Microservices Infrastructure Generator

MicroServices Topology Generator (MSTG) is a new modular and scalable tool that allows to simulate both the application and networking layers of a microservices architecture by deploying the topology on either Docker Compose or Kubernetes.Furthermore, it allows to evaluate the integration of telemetry and monitoring technologies such as Application Performance Monitoring (APM) tools

Iness Ben Guirat (ULB, WP1/WP4) - Traffic Analysis by Adversaries with Partial Visibility

Mixnets, a fundamental privacy enhancing technology in the context of anonymous communication, have been extensively studied in terms of a Global passive Adversary (GPA). However, a more realistic adversary that captures only a portion of the network has not yet been studied due to the complexity of mixnets. In this talk we explain how to evaluate mixnets under a Local Adversary. We provide a framework that models mixnet-based systems and captures different types of Local Adversaries, each modeled based on their goals, prior knowledge, and capabilities. We then use this model to perform traffic analysis using the Metropolis-Hastings algorithm in conjunction with a Bayesian inference engine. We show that depending on the Local Adversary, inferring a probability of certain events can be as accurate as being a Global Adversary.

Abdoul Kader KABORE (Université du Luxembourg) - Automated Test-free Repair of Vulnerable Programs

Most program repair techniques rely on test cases as a key ingredient for driving patch generation and validation. Test cases have been successfully leveraged to automate the repair of many bug classes, but relying on them has hindered progress in repairing vulnerabilities. Their scarcity, exacerbated by the fact that, for vulnerabilities, test cases are also exploits, forms a “sound” barrier that we propose to break in this work. To this end, we introduce Nerve, a novel deep learning-based approach for automating vulnerable software repair. Instead of tests, Nerve leverages the signal in the vulnerability detection and fix suggestions of static analysis security testing (SAST) to learn to repair vulnerable code. Nerve’s learning architecture relies on CodeT5 pre-trained model for source code representation, augmented with a mixed learning objective. This involves, first, the use of triplet loss to build an embedding space that brings each vulnerable code closer to good fixes while keeping it away from incorrect fixes. The second learning objective incorporates cosine similarity into its loss function to align its repair candidates with SAST fix suggestions. We assess Nerve on over 3,000 real-world vulnerable code samples in C/C++, C#, Python, and Java programs. We show that Nerve advances the state of the art in neural vulnerability repair, outperforming VulRepair, the incumbent, by 7.5% points in sequence accuracy, 5% points in CodeBLEU, and 3% points in Normalized Discounted Cumulative Gain (NDCG) for the repair of C/C++ programs. For the other programming languages in the dataset, we provide baselines for future work. Finally, we present a comparison of Nerve with large language model (LLM) using prompts that include different inputs. Our results show that NERVE performed better on some samples compared to LLMs, but also demonstrate that, as with NERVE, SAST results have a positive impact on the performance of these LLM.

Salijona Dyrmishi (Université du Luxembourg) - Enhancing Realism in Adversarial Examples: A Comprehensive Exploration across Tabular Data and Text Domains

Adversarial attacks pose a significant security threat in Machine Learning (ML), employing subtle, invisible perturbations on original examples to craft instances that deceive model decisions. While extensively studied in computer vision and diverse domains such as credit scoring, cybersecurity, cyber-physical systems, and natural language processing, recent findings reveal limitations in traditional adversarial attacks. These approaches often yield examples that lack feasibility, failing to map to real-world objects or adhere to imperceptibility requirements. Consequently, such unrealistic perturbations may not accurately represent genuine threats to ML systems.

Yoann MARQUER (Université du Luxembourg) -Security analysis and testing of CPS software components in the COSMOS project

COSMOS is a H2020 European project focusing on improving DevOps solutions during the development of cyber-physical systems, including reducing security vulnerabilities in software components.
When the source code is available, we leverage an analysis framework to perform a static analysis of the source code to locate potential vulnerabilities, a real-time verification of security specifications during code execution, and a machine learning-based approach to recommend a patch.
When the source code is not available, we rely on metamorphic testing to detect vulnerabilities by comparing outputs of the system.
We provide a catalog of metamorphic relations able to detect various vulnerabilities as well as a framework to reduce the cost of metamorphic security testing.

Matthieu Jimenez (Université du Luxembourg) - A Tale of Vulnerability Prediction Modelling

Vulnerability Prediction Modeling refers to the use of machine learning techniques to identify potentially vulnerable parts of source code, in order to guide the security review process. Over the past two decades, numerous methods have been proposed, evolving in tandem with advancements in machine learning and tailored to address the specific challenges of this domain. This presentation will introduce several of these techniques and discuss key obstacles, including the scarcity of data, the unreliability of ground truth, and the complexity of experimental validation, that hinder their application.

Workshop CyberExcellence Luxembourg