This survey paper provides a comprehensive overview of recent research and development in network security that uses graphs and graph-based data representation and analytics. The paper focuses on the graph-based representation of network traffic records and the application of graph-based analytics in intrusion detection and botnet detection. The paper aims to answer several questions related to graph-based approaches in network security, including the types of graphs used to represent network security data, the approaches used to analyze such graphs, the metrics used for detection and monitoring, and the reproducibility of existing works. The paper presents a survey of graph models used to represent, store, and visualize network security data, a survey of the algorithms and approaches used to analyze such data, and an enumeration of the most important graph features used for network security analytics for monitoring and botnet detection. The paper also discusses the challenges and limitations of using graph-based approaches in network security and identifies potential future research directions. Overall, this survey paper provides a valuable resource for researchers and practitioners in the field of network security who are interested in using graph-based approaches for analyzing and detecting malicious activities in networks.

Mobile applications became the main interaction channel in several domains, such as banking. Consequently, understanding user behaviour on those apps has drawn attention in order to extract business-oriented outcomes. By combining Markov Chain and graph theory techniques, we successfully developed a process to model the app, to extract the click high utility events, to score the interest on those events and cluster the groups of interest. We tested our approach on an European bank dataset with over 3.5 millions of user’s session. By implementing our approach, analysts can gain knowledge of user behaviour in terms of events that are important to the domain.

In many sectors, such as banking, mobile applications became the main interaction channel between customers and institutions. This fast-paced market movement forced an increase in the variety of services offered on the mobile applications. Mobile applications usually exchange information via REST APIs. This range of services with the exposure that APIs imposed to the organizations widen the attack surface, where opportunists can try to take advantages of bad implemented APIs. In light of that, our work focuses in detecting malicious attempts to accessing the APIs by analysing the behaviour of the mobile application user sessions and capturing the anomalies that deviates from the expected. To this end, we tested a pipeline that extracts features based on Markov Chain, Levenshtein Distance and LSTM frequencies to create a machine learning model that was tested in 40 days of data from a retail banking mobile application. By mixing the features from Markov Chain and Levenshtein Distance, our model reached over 96.61% of f-measure. We also present results to discourage the use of LSTM for this use case because it is computationally costly and in fact degraded the performance of the model based on the other frequencies.

The heterogeneous massive logs incoming from multiple sources pose major challenges to professionals responsible for IT security and system administrator. One of the challenges is to develop a scalable heterogeneous logs database for storage and further analysis. In fact, it is difficult to decide which database is suitable for the needs, the best of a use case, execution time and storage performances. In this paper, we explore, study, and compare the performance of SQL and NoSQL databases on large heterogeneous event logs. We implement the relational database using MySQL, the column-oriented database using Impala on the top of Hadoop, and the graph database using Neo4j. We experiment the databases on a large heterogeneous logs and provide advice, the pros and cons of each SQL and NoSQL database. Our findings that Impala outperforms MySQL and Neo4j databases in terms of loading logs, execution time of simple queries, and storage of logs. However, Neo4j outperforms Impala and MySQL in the execution time of complex queries.

Cyber-attacks become more sophisticated and complex especially when adversaries steal user credentials to traverse the network of an organization. Detecting a breach is extremely difficult and this is confirmed by the findings of studies related to cyber-attacks on organizations. A study conducted last year by IBM found that it takes 206 days on average to US companies to detect a data breach. As a consequence, the effectiveness of existing defensive tools is in question. In this work we deal with the detection of malicious authentication events, which are responsible for effective execution of the stealthy attack, called lateral movement. Authentication event logs produce a pure categorical feature space which creates methodological challenges for developing outlier detection algorithms. We propose an auto semi-supervised outlier ensemble detector that does not leverage the ground truth to learn the normal behavior. The automatic nature of our methodology is supported by established unsupervised outlier ensemble theory. We test the performance of our detector on a real-world cyber security dataset provided publicly by the Los Alamos National Lab. Overall, our experiments show that our proposed detector outperforms existing algorithms and produces a 0 False Negative Rate without missing any malicious login event and a False Positive Rate which improves the state-of-the-art. In addition, by detecting malicious authentication events, compared to the majority of the existing works which focus solely on detecting malicious users or computers, we are able to provide insights regarding when and at which systems malicious login events happened. Beyond the application on a public dataset we are working with our industry partner, POST Luxembourg, to employ the proposed detector on their network.

Robot Operating System (ROS) is becoming more and more important and is used widely by developers and researchers in various domains. One of the most important fields where it is being used is the self-driving cars industry. However, this framework is far from being totally secure, and the existing security breaches do not have robust solutions. In this paper we focus on the camera vulnerabilities, as it is often the most important source for the environment discovery and the decision-making process. We propose an unsupervised anomaly detection tool for detecting suspicious frames incoming from camera flows. Our solution is based on spatio-temporal autoencoders used to truthfully reconstruct the camera frames and detect abnormal ones by measuring the difference with the input. We test our approach on a real-word dataset, i.e. flows coming from embedded cameras of self-driving cars. Our solution outperforms the existing works on different scenarios.

Managed security service providers increasingly rely on machine-learning methods to exceed traditional, signature-based threat detection and classification methods. As machine-learning often improves with more data available, smaller organizations and clients find themselves at a disadvantage: Without the ability to share their data and others willing to collaborate, their machine-learned threat detection will perform worse than the same model in a larger organization. We show that Federated Learning, i.e. collaborative learning without data sharing, successfully helps to overcome this problem. Our experiments focus on a common task in cyber security, the detection of unwanted URLs in network traffic seen by security-as-a-service providers. Our experiments show that i) Smaller participants benefit from larger participants ii) Participants seeing different types of malicious traffic can generalize better to unseen types of attacks, increasing performance by 8% to 15% on average, and up to 27% in the extreme case. iii) Participating in Federated training never harms the performance of the locally trained model. In our experiment modeling a security-as-a service setting, Federated Learning increased detection up to 30% for some participants in the scheme. This clearly shows that Federated Learning is a viable approach to address issues of data sharing in common cyber security settings.

•A new graph embedding is proposed for graph streams.•A new incremental similarity measure based on graph edit distance•Dealing with anomaly detection in a stream of heterogeneous labeled graphs.•Allow the detection of anomalies in real-time.

In this paper, we leverage the newly integrated extended Berkely Packet Filters (eBPF) and eXpress Data Path (XDP) to build ROS-FM, a high-performance inline network-monitoring framework for ROS. We extend the framework with a security policy enforcement tool and distributed data visualization tool for ROS1 and ROS2 systems. We compare the overhead of this framework against the generic ROS monitoring tools, and we test the policy enforcement against existing ROS penetration testing tools to evaluate their effectiveness. We find that the network monitoring framework and the associated visualization tools outperform the existing ROS monitoring tools for all robots with more than 10 running processes and that the monitoring tool uses only 4% of the overhead of the generic tools for robots with 80 processes. We further demonstrate the effectiveness of the security tool against common attacks in both ROS1 and ROS2.

A large body of research has been accomplished on prevention and detection of malicious events, attacks, threats, or botnets. However, there is a lack of automatic and sophisticated methods for investigating malicious events/users, understanding the root cause of attacks, and discovering what is really happening before an attack. In this paper, we propose an attack model discovery approach for investigating and mining malicious authentication events across user accounts. The approach is based on process mining techniques on event logs reaching attacks in order to extract the behavior of malicious users. The evaluation is performed on a publicly large dataset, where we extract models of the behavior of malicious users via authentication events. The results are useful for security experts in order to improve defense tools by making them robust and develop attack simulations.

The Domain Name System (DNS), a fundamental protocol that controls how users interact with the Internet, inadequately provides protection for user privacy. Recently, there have been advancements in the field of DNS privacy and security in the form of the DNS over TLS (DoT) and DNS over HTTPS (DoH) protocols. The advent of these protocols and recent advancements in large-scale data processing have drastically altered the threat model for DNS privacy. Users can no longer rely on traditional methods, and must instead take active steps to ensure their privacy. In this paper, we demonstrate how the extended Berkeley Packet Filter (eBPF) can assist users in maintaining their privacy by leveraging eBPF to provide privacy across standard DNS, DoH, and DoT communications. Further, we develop a method that allows users to enforce application-specific DNS servers. Our method provides users with control over their DNS network traffic and privacy without requiring changes to their applications while adding low overhead.

Robotic Operating System(ROS) security research is currently in a preliminary state, with limited research in tools or models. Considering the trend of digitization of robotic systems, this lack of foundational knowledge increases the potential threat posed by security vulnerabilities in ROS. In this article, we present a new tool to assist further security research in ROS, ROSploit. ROSploit is a modular two-pronged offensive tool covering both reconnaissance and exploitation of ROS systems, designed to assist researchers in testing exploits for ROS.

In robotic systems, the physical world is highly coupled with cyberspace.New threats affect cyber-physical systems as they rely on several sensors to perform critical operations.The most sensitive targets are their location systems, where spoofing attacks can force robots to behave incorrectly.In this paper, we propose a novel anomaly detection approach for sensor spoofing attacks, based on an auto-encoder architecture.After initial training, the detection algorithm works directly on the compressed data by computing the reconstruction errors.We focus on spoofing attacks on Light Detection and Ranging (LiDAR) systems.We tested our anomaly detection approach against several types of spoofing attacks comparing four different compression rates for the auto-encoder.Our approach has a 99% True Positive rate and a 10% False Negative rate for the 83% compression rate.However, a compression rate of 41% could handle almost all of the same attacks while using half the data.

The Robot Operating System (ROS) are being deployed for multiple life critical activities such as self-driving cars, drones, and industries. However, the security has been persistently neglected, especially the image flows incoming from camera robots. In this paper, we perform a structured security assessment of robot cameras using ROS. We points out a relevant number of security flaws that can be used to take over the flows incoming from the robot cameras. Furthermore, we propose an intrusion detection system to detect abnormal flows. Our defense approach is based on images comparisons and unsupervised anomaly detection method. We experiment our approach on robot cameras embedded on a self-driving car.

SummaryTCP/UDP port scanning or sweeping is one of the most common technique used by attackers to discover accessible and potentially vulnerable hosts and applications. Although extracting and distinguishing different port scanning strategies is a challenging task, the identification of dependencies among probed ports is primordial for profiling attacker behaviors, with a final goal of better mitigating them. In this paper, we propose an approach that allows to track port scanning behavior patterns among multiple probed ports and identify intrinsic properties of observed group of ports. Our method is fully automated based on graph modeling and data mining techniques, including text mining.It provides to security analysts and operators relevant information about services that are jointly targeted by attackers. This is helpful to assess the strategy of the attacker by understanding the types of applications or environment he or she targets. We applied our method to data collected through a large Internet telescope (or darknet).

In this paper we propose ROS-Defender, a holistic approach to secure robotics systems, which integrates a Security Event Management System (SIEM), an intrusion prevention system (IPS) and a firewall for a robotic system. ROS-Defender combines anomaly detection systems at application (ROS) level and network level, with dynamic policy enforcement points using software defined networking (SDN) to provide protection against a large class of attacks. Although SIEMs, IPS, and firewall have been previously used to secure computer networks, ROSDefender is applying them for the specific use case of robotic systems, where security is in many cases an afterthought.

The advent of massive and highly heterogeneous information systems poses major challenges to professionals responsible for IT security. The huge amount of monitoring data currently being generated means that no human being or group of human beings can cope with their analysis. Furthermore, fully automated tools still lack the ability to track the associated events in a fine-grained and reliable way. Here, we propose the HuMa framework for detailed and reliable analysis of large amounts of data for security purposes. HuMa uses a multi-analysis approach to study complex security events in a large set of logs. It is organized around three layers: the event layer, the context and attack pattern layer, and the assessment layer. We describe the framework components and the set of complementary algorithms for security assessment. We also provide an evaluation of the contribution of the context and attack pattern layer to security investigation.

Anomaly detection on security logs is receiving more and more attention. Authentication events are an important component of security logs, and being able to produce trustful and accurate predictions minimizes the effort of cyber-experts to stop false attacks. Observed events are classified into Normal, for legitimate user behavior, and Malicious, for malevolent actions. These classes are consistently excessively imbalanced which makes the classification problem harder; in the commonly used Los Alamos dataset, the malicious class comprises only 0.00033% of the total. This work proposes a novel method to extract advanced composite features, and a supervised learning technique for classifying authentication logs trustfully; the models are Random Forest, LogitBoost, Logistic Regression, and ultimately Majority Voting which leverages the predictions of the previous models and gives the final prediction for each authentication event. We measure the performance of our experiments by using the False Negative Rate and False Positive Rate. In overall we achieve 0 False Negative Rate (i.e. no attack was missed), and on average a False Positive Rate of 0.0019.