Abstract
Within both the cyber kill chain and MITRE ATT&CK frameworks, Lateral Movement (LM) is defined as any activity that allows adversaries to progressively move deeper into a system in seek of high-value assets. Although this timely subject has been studied in the cybersecurity literature to a significant degree, so far, no work provides a comprehensive survey regarding the identification of LM from mainly an Intrusion Detection System (IDS) viewpoint. To cover this noticeable gap, this work provides a systematic, holistic overview of the topic, not neglecting new communication paradigms, such as the Internet of Things (IoT). The survey part, spanning a time window of eight years and 53 articles, is split into three focus areas, namely, Endpoint Detection and Response (EDR) schemes, machine learning oriented solutions, and graph-based strategies. On top of that, we bring to light interrelations, mapping the progress in this field over time, and offer key observations that may propel LM research forward.

Abstract
Artificial Intelligence (AI) and blockchain technology have emerged as increasingly prevalent and influential elements shaping global trends in Information and Communications Technology (ICT). Namely, the synergistic combination of blockchain and AI introduces beneficial, unique features with the potential to enhance the performance and efficiency of existing ICT systems. However, presently, the confluence of these two disruptive technologies remains in a rather nascent stage, undergoing continuous exploration and study. In this context, the work at hand offers insight regarding the most significant features of the AI and blockchain intersection. Sixteen outstanding, recent articles exploring the combination of AI and blockchain technology have been systematically selected and thoroughly investigated. From them, fourteen key features have been extracted, including data security and privacy, data encryption, data sharing, decentralized intelligent system, efficiency, automated decision system, collective decision-making, scalability, system security, transparency, sustainability, device cooperation, and mining hardware design. Moreover, drawing upon the related literature stemming from major digital databases, we construct a timeline of this technological convergence comprising three eras: emerging, convergence, and application. For the convergence era, we categorize the pertinent features into three primary groups: data manipulation, potential applicability to legacy systems, and hardware issues. For the application era, we elaborate on the impact of this technology fusion from the perspective of five distinct focus areas, from Internet of Things applications and cybersecurity, to finance, energy, and smart cities. This multifaceted but succinct analysis is instrumental in delineating the timeline of AI and blockchain convergence and pinpointing the unique characteristics inherent in their integration. The paper culminates by highlighting the prevailing challenges and unresolved questions in blockchain and AI-based systems, thereby charting potential avenues for future scholarly inquiry.

Abstract
Nowadays, USB is one of the most adopted communication standards. However, the ubiquity of this technology has attracted the interest of attackers. This situation is alarming, considering that the USB protocol has penetrated even into critical infrastructures. Unfortunately, the majority of the contemporary security detection and prevention mechanisms against USB-specific attacks work at the application layer of the USB protocol stack and, therefore, can only provide partial protection, assuming that the host is not itself compromised. Toward this end, we propose a USB authentication system designed to identify (and possibly block) heterogeneous USB-based attacks directly from the physical layer. Empirical observations demonstrate that any extraneous/malicious activity initiated by malicious/compromised USB peripherals tends to consume additional electrical power. Driven by this observation, our proposed solution is based on the analysis of the USB power consumption patterns. Valuable power readings can easily be obtained directly by the power lines of the USB connector with low-cost, off-the-shelf equipment. Our experiments demonstrate the ability to effectively distinguish benign from malicious USB devices, as well as USB peripherals from each other, relying on the power side channel. At the core of our analysis lies an Autoencoder model that handles the feature extraction process; this process is paired with a Long-Short Term Memory (LSTM) and a Convolutional Neural Networks (CNN) model for detecting malicious peripherals. We meticulously evaluated the effectiveness of our approach and compared its effectiveness against various other shallow Machine Learning (ML) methods. The results indicate that the proposed scheme can identify USB devices as benign or malicious/counterfeit with perfect F1-Score.

Abstract
Lateral movement (LM) is an umbrella term for techniques through which attackers spread from an entry point to the rest of the network. Typically, LM involves both pivoting through multiple systems and privilege escalation. As LM techniques proliferate and evolve, there is a need for advanced security controls able to detect and possibly nip such attacks in the bud. Based on the published literature, we argue that although LM-focused intrusion detection systems have received considerable attention, a prominent issue remains largely unaddressed. This concerns the detection of LM through unsupervised machine learning (ML) techniques. This work contributes to this field by capitalizing on the LMD-2023 dataset containing traces of 15 diverse LM attack techniques as they were logged by the system monitor (Sysmon) service of the MS Windows platform. We provide a panorama of this sub-field and associated methodologies, exploring the potential of standard ML-based detection. In further detail, in addition to analyzing feature selection and preprocessing, we detail and evaluate a plethora of unsupervised ML techniques, both shallow and deep. The derived scores for the best performer in terms of the AUC and F1 metrics are quite promising, around 94.7%/93% and 95.2%/93.8%, for the best shallow and deep neural network model, respectively. On top of that, in an effort to further improve on those metrics, we devise and evaluate a two-stage ML model, surpassing the previous best score by approximately 3.5%. Overall, to our knowledge, this work provides the first full-blown study on LM detection via unsupervised learning techniques, therefore it is anticipated to serve as a groundwork for anyone working in this timely field.

Abstract
Lateral movement (LM) is a principal, increasingly common, tactic in the arsenal of advanced persistent threat (APT) groups and other less or more powerful threat actors. It concerns techniques that enable a cyberattacker, after establishing a foothold, to maintain ongoing access and penetrate further into a network in quest of prized booty. This is done by moving through the infiltrated network and gaining elevated privileges using an assortment of tools. Concentrating on the MS Windows platform, this work provides the first to our knowledge holistic methodology supported by an abundance of experimental results towards the detection of LM via supervised machine learning (ML) techniques. We specifically detail feature selection, data preprocessing, and feature importance processes, and elaborate on the configuration of the ML models used. A plethora of ML techniques are assessed, including 10 base estimators, one ensemble meta-estimator, and five deep learning models. Vis-à-vis the relevant literature, and by considering a highly unbalanced dataset and a multiclass classification problem, we report superior scores in terms of the F1 and AUC metrics, 99.41% and 99.84%, respectively. Last but not least, as a side contribution, we offer a publicly available, open-source tool, which can convert Windows system monitor logs to turnkey datasets, ready to be fed into ML models.

Abstract
The impact that IoT technologies have on our everyday life is indisputable. Wearables, smart appliances, lighting, security controls, and others make our life simpler and more comfortable. For the sake of easy monitoring and administration, such devices are typically accompanied by smartphone apps, which are becoming increasingly popular, and sometimes are even required to operate the device. Nevertheless, the use of such apps may indirectly augment the attack surface of the IoT device itself and expose the end-user to security and privacy breaches. Therefore, a key question arises: Do these apps curtail their functionality to the minimum needed, and additionally, are they secure against known vulnerabilities and flaws? In seek of concrete answers to the aforesaid question, this work scrutinizes more than forty chart-topping Android official apps belonging to six diverse mainstream categories of IoT devices. We attentively analyze each app statically, and almost half of them dynamically, after pairing them with real-life IoT devices. The results collected span several axes, namely sensitive permissions, misconfigurations, weaknesses, vulnerabilities, and other issues, including trackers, manifest data, shared software, and more. The short answer to the posed question is that the majority of such apps still remain susceptible to a range of security and privacy issues, which in turn, and at least to a significant degree, reflects the general proclivity in this ecosystem.

Abstract
Wi-Fi is arguably the most proliferated wireless technology today. Due to its massive adoption, Wi-Fi deployments always remain in the epicenter of attackers and evildoers. Surprisingly, research regarding machine learning driven intrusion detection systems (IDS) that are specifically optimized to detect Wi-Fi attacks is lagging behind. On top of that, the field is dominated by false or half-true assumptions that potentially can lead to corresponding models being overfilled to certain validation datasets, simply giving the impression or illusion of high efficiency. This work attempts to provide concrete answers to the following key questions regarding IEEE 802.11 machine learning driven IDS. First, from an expert's viewpoint and with reference to the relevant literature, what are the criteria for determining the smallest possible set of classification features, which are also common and potentially transferable to virtually any deployment types/versions of 802.11? And second, based on these features, what is the detection performance across different network versions and diverse machine learning techniques, i.e., shallow versus deep learning ones? To answer these questions, we rely on the renowned 802.11 security-oriented AWID family of datasets. In a nutshell, our experiments demonstrate that with a rather small set of 16 features and without the use of any optimization or ensemble method, shallow and deep learning classification can achieve an average F1 score of up to 99.55\% and 97.55\%, respectively. We argue that the suggested human expert driven feature selection leads to lightweight, deployment-agnostic detection systems, and therefore can be used as a basis for future work in this interesting and rapidly evolving field.

Abstract
Intrusion detection in wireless and, more specifically, Wi-Fi networks is lately increasingly under the spotlight of the research community. However, the literature currently lacks a comprehensive assessment of the potential to detect application layer attacks based on both 802.11 and non-802.11 network protocol features. The investigation of this capacity is of paramount importance, since Wi-Fi domains are often used as a stepping stone by threat actors for unleashing an ample variety of application layer assaults. In this setting, by exploiting the contemporary AWID3 benchmark dataset along with both shallow and deep learning machine learning techniques, this work attempts to provide concrete answers to a dyad of principal matters. First, what is the competence of 802.11-specific and non-802.11 features when used separately and in tandem in detecting application layer attacks, say, website spoofing? Second, which network protocol features are the most informative to the machine learning model for detecting application layer attacks? Without relying on any optimization or dimensionality reduction technique, our experiments, indicatively exploiting an engineered feature, demonstrate a detection performance up to 96.7% in terms of the Area under the ROC Curve (AUC) metric.

Abstract
This work attempts to answer in a clear way the following key questions regarding the optimal initialization of the Sysmon tool for the identification of Lateral Movement in the MS Windows ecosystem. First, from an expert’s standpoint and with reference to the relevant literature, what are the criteria for determining the possibly optimal initialization features of the Sysmon event monitoring tool, which are also applicable as custom rules within the config.xml configuration file? Second, based on the identified features, how can a functional configuration file, able to identify as many LM variants as possible, be generated? To answer these questions, we relied on the MITRE ATT and CK knowledge base of adversary tactics and techniques and focused on the execution of the nine commonest LM methods. The conducted experiments, performed on a properly configured testbed, suggested a great number of interrelated networking features that were implemented as custom rules in the Sysmon’s config.xml file. Moreover, by capitalizing on the rich corpus of the 870K Sysmon logs collected, we created and evaluated, in terms of TP and FP rates, an extensible Python .evtx file analyzer, dubbed PeX, which can be used towards automatizing the parsing and scrutiny of such voluminous files. Both the .evtx logs dataset and the developed PeX tool are provided publicly for further propelling future research in this interesting and rapidly evolving field.