Cybersecurity exercises are a common way to train and evaluate the skills of cybersecurity professionals. These exercises also provide a unique opportunity to generate datasets with realistic attack traces on non-sensitive systems. Nevertheless, the collected logs are unlabeled, and deciding which logs are related to pentesters is a difficult problem. In this paper, we present a novel methodology to label efficiently both system and network logs using MITRE ATT&CK techniques. To demonstrate the effectiveness of our approach, we introduce CasinoLimit, a dataset generated from a pentest exercise that has been played by 114 participants where we collected 540 GB of attack data. We apply our methodology to accurately label these logs with a semiautomatic approach: labels are inferred from the shell sessions and propagated to the network sessions, and eventually corrected by a junior analyst. An expert analyst has manually reviewed all the labels that have been computed to ensure the quality of the labeling process. The results of the pentest exercise are deeply discussed. We show the variability of players’ behaviors and that players can be distinguished by their command line habits. In addition, the high level of granularity of labels coupled with the number of participants enables multiple other applications. With this paper, we release the full dataset and the associated labeling tool, Manatee, which can be used to browse the logs and labels. To support the generalization of our approach, we made it possible to load other datasets with this tool.

Timed Automata (TA) are formal models capable of representing regular languages with timing constraints, making them well-suited for modeling systems where behavior is driven by events occurring over time. Most existing work on TA learning relies on active learning, where access to a teacher is assumed to answer membership queries and provide counterexamples. While this framework offers strong theoretical guarantees, it is impractical for many real-world applications where such a teacher is unavailable. In contrast, passive learning approaches aim to infer TA solely from sequences accepted by the target automaton. However, current methods struggle to handle noise in the data, such as symbol omissions, insertions, or permutations, often resulting in excessively large and inaccurate automata. In this paper, we introduce TADAM, a novel approach that leverages the Minimum Description Length (MDL) principle to balance model complexity and data fit, allowing it to distinguish between meaningful patterns and noise. We show that TADAM is significantly more robust to noisy data than existing techniques, less prone to overfitting, and produces concise models that can be manually audited. We further demonstrate its practical utility through experiments on real-world tasks, such as network flow classification and anomaly detection.

This paper focuses on graphical models for modelling preferences in combinatorial space and their use for item optimisation. The preferential optimisation task seeks to find the preferred item containing some defined values, which is useful for many recommendation settings in e-commerce. We show that efficient (i.e., with polynomial time complexity) preferential optimisation is achieved with a subset of cyclic CP-nets called conditional acyclic CP-net. We also introduce a new graphical preference model, called Conditional-Optimality networks (CO-networks), that are more concise than conditional acyclic CP-nets and LP-trees but have the same expressiveness with respect to optimisation. Finally, we empirically show that preferential optimisation can be used for encoding alternatives into partial instantiations and vice versa, paving the way towards CO-nets and CP-nets unsupervised learning with the minimal description length (MDL) principle.

The use of Machine Learning for anomaly detection in cyber security-critical applications, such as intrusion detection systems, has been hindered by the lack of explainability. Without understanding the reason behind anomaly alerts, it is too expensive or impossible for human analysts to verify and identify cyber-attacks. Our research addresses this challenge and focuses on unsupervised network intrusion detection, where only benign network traffic is available for training the detection model. We propose a novel post-hoc explanation method, called AE-pvalues, which is based on the p-values of the reconstruction errors produced by an Auto-Encoder-based anomaly detection method. Our work identifies the most informative network traffic features associated with an anomaly alert, providing interpretations for the generated alerts. We conduct an empirical study using a large-scale network intrusion dataset, CICIDS2017, to compare the proposed AE-pvalues method with two state-of-the-art baselines applied in the unsupervised anomaly detection task. Our experimental results show that the AE-pvalues method accurately identifies abnormal influential network traffic features. Furthermore, our study demonstrates that the explanation outputs can help identify different types of network attacks in the detected anomalies, enabling human security analysts to understand the root cause of the anomalies and take prompt action to strengthen security measures.

International audience

Android security has received a lot of attention over the last decade, especially malware investigation. Researchers attempt to highlight applications’ security-relevant characteristics to better understand malware and effectively distinguish malware from benign applications. The accuracy and the completeness of their proposals are evaluated experimentally on malware and goodware datasets. Thus, the quality of these datasets is of critical importance: if the datasets are outdated or not representative of the studied population, the conclusions may be flawed. We specify different types of experimental scenarios. Some of them require unlabeled but representative datasets of the entire population. Others require datasets labeled with valuable characteristics that may be difficult to compute, such as malware datasets. We discuss the irregularities of datasets used in experiments, questioning the validity of the performances reported in the literature. This article focuses on providing guidelines for designing debiased datasets. First, we propose guidelines for building representative datasets from unlabeled ones. Second, we propose and experiment a debiasing algorithm that, given a biased labeled dataset and a target representative dataset, builds a representative and labeled dataset. Finally, from the previous debiased datasets, we produce datasets for experiments on Android malware detection or classification with machine learning algorithms. Experiments show that debiased datasets perfom better when classifying with machine learning algorithms.

The expansion of the Internet-of-Things (IoT) market is visible in homes, factories, public places, and smart cities. While the massive deployment of connected devices offers opportunities to improve quality of life and to develop new services, the impact of such devices on the security of the users in a context where the level of malicious threat continues to increase is a major concern. One of the challenges is the heterogeneity and constant evolution of wireless technologies and protocols used. To overcome this problem, we propose RIDS, a Radio Intrusion Detection System that is based on the monitoring and profiling of radio communications at the physical layer level using autoencoder neural networks. RIDS is independent of the wireless protocols and modulation technologies used. Besides, it is designed to provide a threefold diagnosis of the detected anomalies: temporal (start and end date of the detected anomaly), frequential (main frequency of the anomaly), and spatial (location of the origin of the anomaly). To demonstrate the relevance and the efficiency of our approach, we collected a large dataset of radio-communications recorded with three different probes deployed in an experimental room. Multiple real-world attacks involving a wide variety of communication technologies are also injected to assess the detection and diagnosis efficiency. The results demonstrate the efficiency of RIDS in detecting and diagnosing anomalies that occurred in the 400–500 Mhz and 800–900 Mhz frequency bands. It is noteworthy that compromised devices and attacks using these communication bands are generally not easily covered by traditional solutions.

Nowadays, air travel is one of the safest transportation means. While safety is historically well integrated into avionics systems, it is becoming increasingly important to take into account the security of such systems for the future. In particular, Host-based Intrusion Detection Systems (HIDS) are commonly used in traditional information systems to improve their security. The adaptation of such systems for deployment inside an aircraft has been studied in another work and has shown to be effective in detecting anomalous behavior in an avionic application. However, the detection itself is not sufficient to provide an on-board reaction, and to prevent such intrusion. This paper proposes to improve such HIDS by introducing a signature-based system capable of providing a first diagnosis after the detection of an anomalous behavior. The proposed diagnosis approach is based on the definition of the signature of an alert, and its comparison with a knowledge database that is regularly updated throughout aircraft lifetime. This approach has been implemented on a real avionic computer and yielded good results in terms of classification accuracy and resources consumption.

Energy providers are massively deploying devices to manage distributed resources or equipment. These devices are used for example to manage the energy of smart factories efficiently or to monitor the infrastructure of smart-grids. By design, they typically exhibit homogeneous behavior, with similar software and hardware architecture. Unfortunately, these devices are also of interest to attackers aiming to develop botnets or compromise companies' security. This paper presents a new protection approach based on Hardware Performance Counters (HPC) to detect anomalies in massively deployed devices. These HPC are processed using outlier detection algorithms. Compared to existing solutions, we propose a lightweight approach based on a comparative analysis of devices' HPC without relying on the modeling of the software applications running on the devices. To assess the relevance and the effectiveness of the approach, a thorough experimental analysis is carried out in a representative industrial-type environment, sampling the data from 100 Raspberry Pi to simulate about 10,000 devices deployed simultaneously. The results show high detection and performance efficiency under different software profiles and attack payloads. Moreover, the calibration of the approach depends primarily on the hardware rather than the application software running on the devices. It should ease its deployment in an operational environment.

This paper considers the task of learning the preferences of users on a combinatorial set of alternatives, as it can be the case for example with online configurators. In many settings, what is available to the learner is a set of positive examples of alternatives that have been selected during past interactions. We propose to learn a model of the users' preferences that ranks previously chosen alternatives as high as possible. In this paper, we study the particular task of learning conditional lexicographic preferences. We present an algorithm to learn several classes of lexicographic preference trees, prove convergence properties of the algorithm, and experiment on both synthetic data and on a real-world bench in the domain of recommendation in interactive configuration.