Advanced Attacks Research Articles

Blockchain and Access Control Encryption-Empowered IoT Knowledge Sharing for Cloud-Edge Orchestrated Personalized Privacy-Preserving Federated Learning

Federated learning (FL) is emerging as a powerful paradigm for distributed data mining in the context of Internet of Things (IoT) big data. It addresses privacy concerns associated with data outsourcing by enabling local data training and knowledge (i.e., model) sharing. However, simplistic local knowledge sharing can inadvertently expose user privacy to advanced attacks, such as model inversion or gradient leakage. Furthermore, achieving fine-grained and personalized privacy protection for IoT users remains a challenge. In this paper, we propose a novel solution called hierarchical blockchain-empowered cloud-edge orchestrated federated learning (HBCE-FL) to address these challenges. HBCE-FL is designed to provide secure, intelligent, and distributed data analysis for IoT users. To tackle FL’s privacy issues, we develop a multi-level access control encryption and blockchain-based approach for sharing IoT knowledge within the HBCE-FL framework. Our approach classifies IoT users into different levels based on their individual privacy requirements, enabling fine-grained privacy protection. The blockchain is employed for identity authentication, key management, and message sanitization. For scenarios involving IoT users with non-IID data, we integrate federated multi-task learning into HBCE-FL to ensure fairness, robustness, and privacy. Finally, we conduct experiments using classic MNIST and CIFAR10 datasets to validate our approach. The experimental results illustrate that HBCE-FL effectively achieves personalized privacy-preserving FL without losing IoT data availability. Regardless of whether IoT data are homogeneous or heterogeneous, our approach enhances model accuracy and convergence rates by enabling secure IoT knowledge access and sharing for IoT users.

WaTrojan: Wavelet domain trigger injection for backdoor attacks

Backdoor attacks have been proven to pose effective threats to deep neural networks in various domains, such as biometrics, authentication, and autonomous driving. Attackers compromise the integrity of the model, causing it to behave normally on benign samples under normal circumstances but perform attacker-specified actions on samples containing specific triggers. However, existing attack methods often suffer from two main drawbacks: permissions and concealment. While some attack methods may not require high levels of permissions from the attacker, the triggers are typically visible to the naked eye, significantly reducing the attack's concealment and making it susceptible to detection by existing defense mechanisms. Although many advanced attack methods enhance concealment, they often necessitate control over the model's training process, thereby significantly limiting the practical applicability of the attack. To circumvent the two aforementioned drawbacks, we propose a novel backdoor attack method called WaTrojan, which implements the attack by adding triggers in the wavelet domain. The key to this attack lies in adding perturbations to the wavelet domain of an image, thereby altering the entire spatial domain of pixels. This approach challenges many assumptions of existing defense methods and makes poisoned images nearly indistinguishable from clean images visually. We evaluate WaTrojan on five benchmark datasets, including MNIST, CIFAR-10, GTSRB, CelebA, and ImageNet. The results indicate that our attack achieves an extremely high attack success rate while causing almost no drop in accuracy on benign samples. The visual quality of the poisoned images is high, with little perceptual difference from benign images. Furthermore, we assess the performance of WaTrojan under existing defense measures, and the results show that WaTrojan is robust and can significantly evade and resist the impacts generated by these defense measures.

Clustering-based attack detection for adversarial reinforcement learning

Detecting malicious attacks presents a major challenge in the field of reinforcement learning (RL), as such attacks can force the victim to perform abnormal actions, with potentially severe consequences. To mitigate these risks, current research focuses on the enhancement of RL algorithms with efficient detection mechanisms, especially for real-world applications. Adversarial attacks have the potential to alter the environmental dynamics of a Markov Decision Process (MDP) perceived by an RL agent. Leveraging these changes in dynamics, we propose a novel approach to detect attacks. Our contribution can be summarized in two main aspects. Firstly, we propose a novel formalization of the attack detection problem that entails analyzing modifications made by attacks to the transition and reward dynamics within the environment. This problem can be framed as a context change detection problem, where the goal is to identify the transition from a “free-of-attack” situation to an “under-attack” scenario. To solve this problem, we propose a groundbreaking “model-free” clustering-based countermeasure. This approach consists of two essential steps: first, partitioning the transition space into clusters, and then using this partitioning to identify changes in environmental dynamics caused by adversarial attacks. To assess the efficiency of our detection method, we performed experiments on four established RL domains (grid-world, mountain car, carpole, and acrobot) and subjected them to four advanced attack types. Uniform, Strategically-timed, Q-value, and Multi-objective. Our study proves that our technique has a high potential for perturbation detection, even in scenarios where attackers employ more sophisticated strategies.

Designing energy-aware collaborative intrusion detection in IoT networks

The Internet of Things (IoT) with its evolution brings many benefits to people’s routine life, while at the same time posing various security challenges, due to the lack of software updates and access control policies. For protection, collaborative/distributed detection networks are one essential security solution, which can enhance the detection capability of a separate detection node by governing a set of eligible detection nodes to share information. However, internal intruders/attacks are the main threat source to the distributed structure including both IoT and CIDNs. In the literature, the CIDNs with challenge-based trust mechanism is found to be secure under many common external threats, but would be ineffective to advanced insider threats. Motivated by this issue, this work considers the factor of energy consumption and develops an enhanced challenge-based CIDN, called EnergyCIDN. The main idea relies on the observation that advanced attacks should consume more energy than a benign node. Under the evaluation on our EnergyCIDN, we conduct a set of experiments and tests in many scenarios, including simulated network, practical IoT network, and medical network. In comparison with the existing similar approaches, our EnergyCIDN showcases better detection capability regarding both common and advanced insider attacks.

A comprehensive guide to CAN IDS data and introduction of the ROAD dataset.

Although ubiquitous in modern vehicles, Controller Area Networks (CANs) lack basic security properties and are easily exploitable. A rapidly growing field of CAN security research has emerged that seeks to detect intrusions or anomalies on CANs. Producing vehicular CAN data with a variety of intrusions is a difficult task for most researchers as it requires expensive assets and deep expertise. To illuminate this task, we introduce the first comprehensive guide to the existing open CAN intrusion detection system (IDS) datasets. We categorize attacks on CANs including fabrication (adding frames, e.g., flooding or targeting and ID), suspension (removing an ID's frames), and masquerade attacks (spoofed frames sent in lieu of suspended ones). We provide a quality analysis of each dataset; an enumeration of each datasets' attacks, benefits, and drawbacks; categorization as real vs. simulated CAN data and real vs. simulated attacks; whether the data is raw CAN data or signal-translated; number of vehicles/CANs; quantity in terms of time; and finally a suggested use case of each dataset. State-of-the-art public CAN IDS datasets are limited to real fabrication (simple message injection) attacks and simulated attacks often in synthetic data, lacking fidelity. In general, the physical effects of attacks on the vehicle are not verified in the available datasets. Only one dataset provides signal-translated data but is missing a corresponding "raw" binary version. This issue pigeon-holes CAN IDS research into testing on limited and often inappropriate data (usually with attacks that are too easily detectable to truly test the method). The scarcity of appropriate data has stymied comparability and reproducibility of results for researchers. As our primary contribution, we present the Real ORNL Automotive Dynamometer (ROAD) CAN IDS dataset, consisting of over 3.5 hours of one vehicle's CAN data. ROAD contains ambient data recorded during a diverse set of activities, and attacks of increasing stealth with multiple variants and instances of real (i.e. non-simulated) fuzzing, fabrication, unique advanced attacks, and simulated masquerade attacks. To facilitate a benchmark for CAN IDS methods that require signal-translated inputs, we also provide the signal time series format for many of the CAN captures. Our contributions aim to facilitate appropriate benchmarking and needed comparability in the CAN IDS research field.

Data-Driven Cyber Threat Intelligence: Leveraging Behavioral Analytics for Proactive Defense Mechanisms

As the cyber threat landscape becomes increasingly sophisticated, traditional defense mechanisms often fall short in anticipating and mitigating advanced attacks. The paper explores the critical role of behavioral analytics in transforming cyber threat intelligence (CTI) into a proactive defense strategy. This study underscores the importance of analyzing user behavior patterns, network activity, and system interactions to detect anomalies that may indicate potential threats. By leveraging large datasets and advanced analytical techniques, organizations can move beyond reactive cybersecurity measures, instead anticipating and preventing attacks before they fully manifest. The integration of behavioral analytics with CTI provides a comprehensive understanding of both external threats and internal vulnerabilities, enabling the development of dynamic defense mechanisms that adapt to the evolving threat environment. The research also discusses the benefits of automated threat intelligence platforms, which use machine learning algorithms to continuously analyze behavioral data and refine threat detection models in real-time. This approach reduces the reliance on manual analysis, enhances the speed and accuracy of threat detection, and minimizes false positives. Moreover, the study highlights the importance of cross-sector collaboration and data sharing in building a robust CTI framework that leverages collective intelligence to combat cyber threats. Case studies of successful implementations in various industries demonstrate the practical applications and efficacy of behavioral analytics in enhancing cybersecurity resilience. The findings suggest that organizations that adopt data-driven, behavior-focused CTI are better equipped to defend against both known and unknown threats, ensuring a more secure digital environment. This proactive approach to cybersecurity not only strengthens organizational defenses but also contributes to the broader goal of creating a safer cyber ecosystem by identifying and neutralizing threats at their inception. Keywords: Data-Driven, Cyber Threat, Intelligence, Bahavioral Analytics, Proactive Defense Mechanism.

Adv-4-Adv: Thwarting changing adversarial perturbations via adversarial domain adaptation

Whereas adversarial training can be useful against specific adversarial perturbations, they have also been proven ineffective in generalizing towards attacks deviating from those used for training. However, we observe that this ineffectiveness is intrinsically connected to domain adaptability, another crucial issue in deep learning for which adversarial domain adaptation appears to be a promising solution. Consequently, we propose Adv-4-Adv as a novel adversarial training method that aims to retain robustness against unseen adversarial perturbations. Essentially, Adv-4-Adv treats attacks incurring different perturbations as distinct domains, and by leveraging the power of adversarial domain adaptation, it aims to remove the domain/attack-specific features. This forces a trained model to learn a robust domain-invariant representation, which in turn enhances its generalizability. Extensive evaluations on Fashion-MNIST, SVHN, CIFAR-10, and CIFAR-100 demonstrate that a model trained by Adv-4-Adv based on examples crafted by simple attacks (e.g., FGSM) can be generalized to more advanced attacks (e.g., PGD), and the performance exceeds state-of-the-art proposals on these datasets.

Improvement Attack Detection on Internet of Thinks Using Principal Component Analysis and Random Forest

Network security has become crucial in facing increasingly complex and sophisticated attack threats. Network intrusion detection aids in identifying suspicious activities indicating unauthorized intrusions. This research aims to enhance the performance of advanced attack detection. The Random Forest method is an algorithm that leverages an ensemble of decision trees. This ensemble comprises several independent decision trees used to classify data. One characteristic of the Random Forest method is its ability to address overfitting issues and provide good predictive quality. One approach to improving RF's performance is through Principal Component Analysis (PCA). PCA is a statistical technique used to reduce feature dimensionality. PCA eliminates feature correlations and identifies essential features that can enhance the detection of attacks and normal traffic. This research will be tested with the CIC IoT 2023 dataset, encompassing various attack types. The model testing consists of four feature dimensions, namely 5, 8, 10, and 47. The detection results are promising, significantly improving attack detection performance, reaching up to 99.2%.

Defense Against On-Chip Trojans Enabling Traffic Analysis Attacks Based on Machine Learning and Data Augmentation

Modern computing systems involve huge data exchange across various sections of the processing system. To facilitate this, Network-on-Chip (NoC) serve as a crucial infrastructure that connect the processing cores to memory, peripherals, etc. The system could be put at great risk should the NoC system become compromised. The NoCs are used in multi/many-core processors; this domain is experiencing increased threats because of Hardware Trojan (HT) embedded in the multi core processing systems due to the presence of third-party entities in the system-on-chip (SoC) design pipeline. Protecting user and system level privacy becomes important in such multi core systems to enable trust. By embedding a hardware Trojan (HT) in a NoC, the adversary can snoop on important insights regarding the applications executing on the system or the user profile information. An attack of such calibre can compromise privacy thereby enabling more advanced attack on the entire system. This work demonstrates the capability of a traffic analysis attack when a few HTs are embedded in the NoC switches of a multi/many-core processor. The attack is capable of exposing sensitive information to an external malicious attacker who can then analyze the payload data with sophisticated machine learning (ML) techniques to infer the applications executing on the system. We also evaluate the performance of a generative adversarial network (GAN) strengthened attacker model that offers more robustness for data paucity scenarios. We propose a Simulated Annealing-based randomized routing algorithm based defense for NoCs thus thwarting the attack. The results demonstrate that the proposed randomized routing algorithm could reduce the accuracy of identifying user profiles by the attacker from >98% to <15% in multi/many-core systems.

Spoofing Attacks and Anti-Spoofing Methods for Face Authentication Over Smartphones

The ubiquitous usages of smartphones demand reliable mobile authentication systems, which are expected to play an indispensable role in protecting users' sensitive and private information. Among various mobile authentication mechanisms, face authentication (FA) has become the most popular choice on modern smartphones, owing to its excellent performance and user experience. Unfortunately, emerging advanced attacks have brought in severe security concerns against FA systems (e.g., privacy leakage, property damage), which have also triggered increased attention on face anti-spoofing designs. This paper provides a concise review of the advanced face anti-spoofing systems over smartphones from a practical perspective. We first review the presentation attacks (PAs) toward FA systems and classify them according to the geometry and dynamic property of the presentation mediums. Then we survey the state-ofthe- art presentation attack detection (PAD) approaches over smartphones and categorize these methods into four categories based on the cues they leverage. Furthermore, we present a comprehensive comparison of their advantages and disadvantages from various practical aspects. Finally, we provide recommendations for future directions. We hope our viewpoints would shed light on potential pathways for more reliable FA in the future.

Collaborative Federated Learning-Based Model for Alert Correlation and Attack Scenario Recognition

Planned and targeted attacks, such as the advanced persistent threat (APT), are highly sophisticated forms of attack. They involve numerous steps and are intended to remain within a system for an extended length of period before progressing to the next stage of action. Anticipating the next behaviors of attackers is a challenging and crucial task due to the stealthy nature of advanced attack scenarios, in addition to the possible high volumes of false positive alerts generated by different security tools such as intrusion detection systems (IDSs). Intelligent models that are capable of establishing a correlation individual between individual security alerts in order to reconstruct attack scenarios and to extract a holistic view of intrusion activities are required to exploit hidden links between different attack stages. Federated learning models performed in distributed settings have achieved successful and reliable implementations. Alerts from distributed security devices can be utilized in a collaborative manner based on several learning models to construct a federated model. Therefore, we propose an intelligent detection system that employs federated learning models to identify advanced attack scenarios such as APT. Features extracted from alerts are preprocessed and engineered to produce a model with high accuracy and fewer false positives. We conducted training on four machine learning models in a centralized learning; these models are XGBoost, Random Forest, CatBoost, and an ensemble learning model. To maintain privacy and ensure the integrity of the global model, the proposed model has been implemented using conventional neural network federated learning (CNN_FL) across several clients during the process of updating weights. The experimental findings indicate that ensemble learning achieved the highest accuracy of 88.15% in the context of centralized learning. CNN_FL has demonstrated an accuracy of 90.18% in detecting various attacks of APTs while maintaining a low false alarm rate.

APTSHIELD: A Stable, Efficient and Real-Time APT Detection System for Linux Hosts

Advanced Persistent Threat (APT) attack usually refers to the form of long-term, covert and sustained attack on specific targets, with an adversary using advanced attack techniques to destroy the key facilities of an organization. APT attacks have caused serious security threats and massive financial loss worldwide. Academics and industry thereby have proposed a series of solutions to detect APT attacks, such as dynamic/static code analysis, traffic detection, sandbox technology, endpoint detection and response (EDR), etc. However, existing defenses are failed to accurately and effectively defend against the current APT attacks that exhibit strong persistent, stealthy, diverse and dynamic characteristics due to the weak data source integrity, large data processing overhead and poor real-time performance in the process of real-world scenarios. To overcome these difficulties, in this paper we propose APTSHIELD, a stable, efficient and real-time APT detection system for Linux hosts. In the aspect of data collection, audit is selected to stably collect kernel data of the operating system so as to carry out a complete portrait of the attack based on comprehensive analysis and comparison of existing logging tools; In the aspect of data processing, redundant semantics skipping and non-viable node pruning are adopted to reduce the amount of data, so as to reduce the overhead of the detection system; In the aspect of attack detection, an APT attack detection framework based on ATT&CK model is designed to carry out real-time attack response and alarm through the transfer and aggregation of labels. Experimental results on both laboratory and Darpa Engagement show that our system can effectively detect web vulnerability attacks, file-less attacks and remote access trojan attacks, and has a low false positive rate, which adds far more value than the existing frontier work.

Host-based intrusion detection with multi-datasource and deep learning

Modern hackers display increasing sophistication. Intrusion detection systems, both network-based and host-based, now utilize machine learning for improved detection of such advanced attacks. While most of these systems rely on a single data source for training, practical scenarios often involve attack features scattered across multiple sources, posing challenges to the system’s effectiveness in detection. This impairs their potential for attack detection. Thus, this study assesses three host-based data sources—network traffic, system logs, and host statistics. It evaluates and compares their combined detection capabilities across diverse attack stages and types. In the proposed framework, network traffic data is handled by a Convolutional Neural Network (CNN) for improved automatic feature selection. System log data are processed using Long Short-Term Memory (LSTM) and an attention model to enhance temporal relationship exploration. Host statistics are processed by a Deep Neural Network (DNN) to improve classification performance. Experimental results show that the F1-scores reach 1.0 for all considered attacks and attack stages when all three data sources are utilized in the detection process. Additionally, employing diverse models based on the data type leads to improved results, a fact exemplified by Lin et al. (2022) which exclusively utilized XGBoost. The host statistics were found to be highly effective in detecting attacks and were thus investigated further for different attack methods and attack stages. The results showed that the disk usage percentage (DSK), minor memory faults (MINFLT), major memory faults (MAJFLT), total virtual memory growth during the last interval (VGROW), and total resident memory growth during the last interval (RGROW) were primarily affected by all the attacks in the initial access and command and control stages. By contrast, in the impact attack stage, the affected system resources varied widely depending on the particular attack.

A novel attention-based feature learning and optimal deep learning approach for network intrusion detection

Rapid technological advances and network progress has occurred in recent decades, as has the global growth of services via the Internet. Consequently, piracy has become more prevalent, and many modern systems have been infiltrated, making it vital to build information security tools to identify new threats. An intrusion detection system (IDS) is a critical information security technology that detects network fluctuations with the help of machine learning (ML) and deep learning (DL) approaches. However, conventional techniques could be more effective in dealing with advanced attacks. So, this paper proposes an efficient DL approach for network intrusion detection (NID) using an optimal weight-based deep neural network (OWDNN). The network traffic data was initially collected from three openly available datasets: NSL-KDD, CSE-CIC-IDS2018 and UNSW-NB15. Then preprocessing was carried out on the collected data based on missing values imputation, one-hot encoding, and normalization. After that, the data under-sampling process is performed using the butterfly-optimized k-means clustering (BOKMC) algorithm to balance the unbalanced dataset. The relevant features from the balanced dataset are selected using inception version 3 with multi-head attention (IV3MHA) mechanism to reduce the computation burden of the classifier. After that, the dimensionality of the selected feature is reduced based on principal component analysis (PCA). Finally, the classification is done using OWDNN, which classifies the network traffic as normal and anomalous. Experiments on NSL-KDD, CSE-CIC-IDS2018 and UNSW-NB15 datasets show that the OWDNN performs better than the other ID methods.

Detecting malicious reviews and users affecting social reviewing systems: A survey

The proliferation of attacks on On-line Social Networks (OSNs) has imposed particular attention by providers and users. This has an even higher importance for Social Reviewing Systems (SRSs), where users can be strongly conditioned by means of malevolent reviews and behavior of fake or camouflage accounts. There is a rich literature on the use of strong authentication means, encryption or privacy-preserving schemes for OSNs and SRSs, but these mechanisms only represent a first defense line. Advanced attacks may be able to bypass such a defense and to successfully threaten the system and harm its users. Therefore, it is needed to embed a second defence line to peculiar attack scenarios of SRSs that take advantage of user dynamics occurring within social networks. This survey focuses on the issue and solutions to detect malicious reviews and users so as to exclude them from social networks, protect legitimate and honest users and keep the credibility of the protected SRS as high as possible. Therefore, we provide an evaluation of the main detection solutions for the mentioned specific attacks against SRSs in according to different metrics on several standard datasets.

A Novel Authentication Method That Combines Honeytokens and Google Authenticator

Despite the rapid development of technology, computer systems still rely heavily on passwords for security, which can be problematic. Although multi-factor authentication has been introduced, it is not completely effective against more advanced attacks. To address this, this study proposes a new two-factor authentication method that uses honeytokens. Honeytokens and Google Authenticator are combined to create a stronger authentication process. The proposed approach aims to provide additional layers of security and protection to computer systems, increasing their overall security beyond what is currently provided by single-password or standard two-factor authentication methods. The key difference is that the proposed system resembles a two-factor authentication but, in reality, works like a multi-factor authentication system. Multi-factor authentication (MFA) is a security technique that verifies a user’s identity by requiring multiple credentials from distinct categories. These typically include knowledge factors (something the user knows, such as a password or PIN), possession factors (something the user has, such as a mobile phone or security token), and inherence factors (something the user is, such as a biometric characteristic like a fingerprint). This multi-tiered approach significantly enhances protection against potential attacks. We examined and evaluated our system’s robustness against various types of attacks. From the user’s side, the system is as friendly as a two-factor authentication method with an authenticator and is more secure.

Multiple Cooperative Attackers for Tag-Based Physical Layer Authentication

This article concerns the problem of ensuring the security of the tag-based Physical-Layer Authentication (PLA) schemes under multiple cooperative attackers. This problem is important and challenging because of the following reasons. First, if more attackers launch a cooperative attack, the success probability of launching an eavesdropping attack will be improved. Second, based on a cooperative communication system, cooperative attackers can launch more advanced attacks, e.g., a cooperative attack that traps the legitimate user to re-transmit the same tag again. Thus, it is challenging and desirable to construct a secure tag-based PLA scheme under multiple cooperative attackers. For effectively defending against a cooperative attack under multiple cooperative attackers, we propose a new tag-based PLA scheme, referred to as the PLA with the Detection of a Cooperative Attack (PLA-DCA) scheme. The basic idea of the PLA-DCA scheme is to utilize the extra noise in a replayed signal, since a replayed signal inevitably changes the distribution property of receiver noise, e.g., increment of the noise variance. Then, we implemented the proposed scheme and conducted extensive performance comparisons to demonstrate the superiority of the proposed scheme. We evaluate the security of various schemes through equivocation, which represents the uncertainty of estimating a tag. From the simulation results, we verify that the proposed scheme provides higher security than the prior tag-based PLA schemes.

Deep-learning-based Intrusion Detection for Software-defined Networking Space Systems

This paper briefly reviews the application of the Software-defined Networking (SDN) architecture to satellite networks. It highlights the prominent cyber threats that SDN-based satellite networks are vulnerable to and proposes relevant defence mechanisms. SDN transforms traditional networking architectures by separating the control plane from the forwarding (data) plane. This separation enhances scalability and centralises management. In comparison, in traditional networks, the control plane and the data plane are usually combined, resulting in complex network management and reduced scalability. Satellite networks can take advantage of these benefits offered by SDN and this supports them as key enablers of critical services, including weather prediction, global broadband Internet coverage, and Internet of Things (IoT) services. Ease of configuration and flexibility are essential for satellites providing critical services to instantly adapt to network changes. These desirable attributes can be realised by applying SDN to satellite networks. Although SDN offers significant benefits to satellite networks, it is vulnerable to cyber-attacks and particularly due to its centralised architecture. A common attack on SDN is the Distributed Denial of Service (DDoS) attack which could render the entire SDN unavailable. To mitigate such threats, an efficient Intrusion Detection System (IDS) is required to monitor the network and detect any suspicious traffic. However, traditional IDSs produce too many false positives and often fail to detect advanced attacks. For their ability to learn feature hierarchies in network traffic data automatically, whether, for network traffic classification or anomaly detection, deep learning (DL) plays an increasingly important role in IDSs. In this paper, we present a brief review of recent developments in cyber security for SDN-based space systems, and we identify vulnerabilities and threats to an SDN-based satellite network. We further discuss the potential of a DL-based IDS for the detection of cyber threats. Finally, we identify further research gaps in the recent literature and propose future research directions.

TTANAD: Test-Time Augmentation for Network Anomaly Detection.

Machine learning-based Network Intrusion Detection Systems (NIDS) are designed to protect networks by identifying anomalous behaviors or improper uses. In recent years, advanced attacks, such as those mimicking legitimate traffic, have been developed to avoid alerting such systems. Previous works mainly focused on improving the anomaly detector itself, whereas in this paper, we introduce a novel method, Test-Time Augmentation for Network Anomaly Detection (TTANAD), which utilizes test-time augmentation to enhance anomaly detection from the data side. TTANAD leverages the temporal characteristics of traffic data and produces temporal test-time augmentations on the monitored traffic data. This method aims to create additional points of view when examining network traffic during inference, making it suitable for a variety of anomaly detector algorithms. Our experimental results demonstrate that TTANAD outperforms the baseline in all benchmark datasets and with all examined anomaly detection algorithms, according to the Area Under the Receiver Operating Characteristic (AUC) metric.

PD-CPS: A practical scheme for detecting covert port scans in high-speed networks

Port scanning is a commonly used technique to discover the vulnerabilities of network systems. Since many existing methods can detect fast port scans effectively, some advanced attackers perform slow port scans to avoid suspicion. Highly covert slow scans can last for tens of days, which poses a great challenge to current intrusion detection methods. In addition, the existing port scanning detection methods are all based on full traffic. They are difficult to apply to high-speed networks due to the huge consumption of computing and storage resources. This paper proposes PD-CPS, which is a Practical scheme for Detecting Covert Port Scans in high-speed networks. Based on the protocol characteristics and the connection mode of port scanning, we construct a traffic feature set that can not only distinguish the specific scan types but also remain valid for the sampled traffic. Furthermore, we customize a data structure Scan Detection Sketch (SDS) for rapid feature extraction. Experimental results using real-world traces demonstrate the feasibility of PD-CPS in detection accuracy and resource consumption. And the proposed method still works well for slow port scans that last for over 60 days.