Attack Methods Research Articles

Improve the effectiveness of machine learning models in detecting website phishing using morphological features in URL analysis

With the proliferation of the Internet, the emergence of various threats has become increasingly prevalent, particularly the danger posed by phishing websites. These websites are designed with malicious content aimed at exploiting users who inadvertently access them. This method of attack represents a significant potential risk for users in cyberspace. The problem of detecting and eliminating phishing websites has garnered significant interest and research within the community. In this study, we propose a set of morphological features in URL path analysis, combined with machine learning methods, to detect phishing website URLs. Experimental evaluation with the UCI Repository dataset results have demonstrated the effectiveness of the proposed feature set in terms of all metrics (Accuracy, Precision, Recall, and F1 Score) compared to previous methods.

LGTDA: Bandwidth exhaustion attack on Ethereum via dust transactions

Dust attacks typically involve sending a large number of low-value transactions to numerous addresses, aiming to facilitate transaction tracking and undermine privacy, while simultaneously disrupting the market and increasing transaction delays. These transactions not only impact the network but also incur significant costs. This paper introduces a low-cost attack method called LGTDA, which achieves network partitioning through dust attacks. This method hinders block synchronization by consuming node bandwidth, leading to denial of service(DoS) for nodes and eventually causing large-scale network partitioning. In LGTDA, the attacker does not need to have real control over the nodes in the network, nor is there a requirement for the number of peer connections to the nodes; the attack can even be initiated by simply invoking RPC services to send transactions. Under the condition of ensuring the validity of the attack transactions, the LGTDA attack sends a large volume of low-value, high-frequency dust transactions to the network, relying on nodes for global broadcasting. This sustained attack can significantly impede the growth of block heights among nodes, resulting in network partitioning. We discuss the implications of the LGTDA attack, including its destructive capability, low cost, and ease of execution. Additionally, we analyze the limitations of this attack. Compared to grid lighting attacks, the LGTDA attack has a broader impact range and is not limited by the positional relationship with the victim node. Through experimental validation in a controlled environment, we confirm the effectiveness of this attack.

Resilient event-triggered [formula omitted] control for autonomous vehicle under novel important-data-based DoS attack

This paper presents the method of a resilient event-triggered H∞ controller for autonomous vehicles (AVs) to address the emerging threat of important-data-based (IDB) Denial-of-Service (DoS) attack. Specifically, a unique IDB DoS attack strategy from an adversarial perspective is proposed to enhance the attack’s destructiveness. Unlike many mature DoS attack methods, the proposed IDB DoS attack could discern the importance level for the data and selectively launch attacks on crucial information, resulting in more significant damage. Furthermore, a novel resilient event-triggered H∞ control technique is employed to mitigate the impact of the attack and counteract the negative effects produced by the IDB DoS attack. Utilizing the Lyapunov functional method, sufficient conditions are successfully derived to ensure the asymptotical stability of the resulted closed-loop system and H∞ interference suppression performance. Subsequently, simulation results demonstrate that: (1) the addressed IDB DoS attack method leads to a more severe system degradation, highlighting the potential risks associated with this novel attack approach; and (2) the proposed event-triggered H∞ controller effectively mitigates the potential impact of the IDB DoS attack, showcasing its efficiency in preserving system performance under adversarial conditions.

General secure encryption algorithm for separable reversible data hiding in encrypted domain

The separable reversible data hiding in encrypted domain (RDH-ED) algorithm leaves out the embedding space for the information before or after encryption and makes the operation of extracting the information and restoring the image not interfere with each other. The encryption method employed not only affects the embedding space of the information and separability, but is more crucial for ensuring security. However, the commonly used XOR, scram-bling or combination methods fall short in security, especially against known plaintext attack (KPA). Therefore, in or-der to improve the security of RDH-ED and be widely applicable, this paper proposes a high-security RDH-ED encryption algorithm that can be used to reserve space before encryption (RSBE) and free space after encryption (FSAE). During encryption, the image undergoes block XOR, global intra-block bit-plane scrambling (GIBS) and inter-block scrambling sequentially. The GIBS key is created through chaotic mapping transformation. Subsequently, two RDH-ED algorithms based on this encryption are proposed. Experimental results indicate that the algorithm outlined in this paper maintains consistent key communication traffic post key conversion. Additionally, its computational complexity remains at a constant level, satisfying separability criteria, and is suitable for both RSBE and FSAE methods. Simultaneously, while satisfying the security of a single encryption technique, we have expanded the key space to 28Np×Np!×8!Np, enabling resilience against various existing attack methods. Notably, particularly in KPA testing scenarios, the average decryption success rate is a mere 0.0067% and 0.0045%, highlighting its exceptional security. Overall, this virtually unbreakable system significantly enhances image security while preserving an appropriate embedding capacity.

A threat modeling framework for IoT-Based botnet attacks

Internet of Things (IoT) devices are much closer to users than personal computers used in traditional computing environments. Due to prevalence of IoT devices, even if they are compromised and used in attacks, it is difficult to detect and respond to them. Currently, there has been extensive research on threat modeling for cyberattacks. However, there remains a significant gap in research concerning threat modeling for attacks specially targeting IoT devices within the fifth-generation communication environment. In this paper, we present IoT Targeting-Threat Modeling(I3TM) framework established by analyzing botnets that are appeared before 2021 such as Mirai, Pink etc. Through this framework, we identify tactics and techniques to respond to the attacks. Using the identified tactics and techniques from our proposed framework, we can promptly respond to the newly detected attacks. We constructed a Threat Modeling Framework Keyword-Based Metrics to show extracted keywords from reports, academic papers, and white paper that identifies the features of botnet. We also provide an objective way to apply those keywords to the framework. Our framework is organized to analyze the attack process of botnets that may occur against IoT. The framework derives execution for each tactic for objective analysis based on keywords. In the validation for the framework, I3TM identified eight Tactics from Medusa botnet. If the application of the I3TM framework is continuously accumulated, a baseline of similar attack methods and data will be formed. In future research, we are planning to append mitigations for the attacks targeting IoT to the I3TM framework.

H3NI: Non-target-specific node injection attacks on hypergraph neural networks via genetic algorithm

Node injection attack is widely used in graph neural networks (GNNs) attacks, which misleads GNNs by injecting nodes. Though hypergraph neural networks (HNNs) are an extension of GNNs, node injection attacks have not yet been studied in HNNs. Since each edge of a hypergraph can connect more than two nodes, existing node injection methods designed for GNNs cannot effectively select the hyperedges connected to the injected nodes when applied to hypergraphs. In this paper, we propose a Hypergraph Neural Network Node Injection attack method called H3NI, which utilizes a genetic algorithm and a predefined budget model to implement the first black-box node injection framework designed for HNNs attacks. We conducted experiments on the datasets of Cora co-authorship and co-citation. Experimental results show the effectiveness and superior performance of H3NI in attacking HNNs, which reduces the model accuracy by 18%–20% within 5% of the total injected nodes and achieves a 2-4X improvement compared to existing gradient-based node injection methods.

Mechanism and Performance Control Methods of Sulfate Attack on Concrete: A Review

For concrete structures in marine or groundwater environments, sulfate attack is a major factor contributing to the degradation of concrete performance. This paper analyzes the existing literature on the chemical reactions and physical crystallization effects of sulfate attack on cement-based materials, summarizing the degradation mechanisms of corroded concrete. Experiments have been conducted to study the performance evolution of concrete under sulfate attack, considering both external environmental factors and internal factors of the cement-based materials. External environmental factors, such as the temperature, humidity, concentration, and type of sulfate solutions, wet-dry cycles, freeze-thaw cycles, chloride coupling effects, and stray currents significantly impact sulfate attack on concrete. Internal factors, including internal sources of corrosion, the chemical composition of the cement, water-cement ratio, and the content of C-S-H gel and Ca(OH)2, influence the density and sulfate resistance of the cement-based materials. Additionally, five typical methods for enhancing the sulfate resistance of concrete are summarized. Finally, the paper identifies current challenges in the study of corroded concrete and proposes directions for future research.

Black-box adversarial attacks against image quality assessment models

The problem of No-Reference Image Quality Assessment (NR-IQA) is to predict the perceptual quality of an image in line with its subjective evaluation. However, the vulnerabilities of NR-IQA models to the adversarial attacks have not been thoroughly studied for model refinement. This paper aims to investigate the potential loopholes of NR-IQA models via black-box adversarial attacks. Specifically, we first formulate the attack problem as maximizing the deviation between the estimated quality scores of original and perturbed images, while restricting the perturbed image distortions for visual quality preservation. Under such formulation, we then design a Bi-directional loss function to mislead the estimated quality scores of adversarial examples towards an opposite direction with maximum deviation. On this basis, we finally develop an efficient and effective black-box attack method for NR-IQA models based on a random search paradigm. Comprehensive experiments on three benchmark datasets show that all evaluated NR-IQA models are significantly vulnerable to the proposed attack method. After being attacked, the average change rates in terms of two well-known IQA performance metrics achieved by victim models reach 97% and 101%, respectively. In addition, our attack method also outperforms a newly introduced black-box attack approach on IQA models. We also observe that the generated perturbations are not transferable, which points out a new research direction in NR-IQA community. The source code is available at https://github.com/GZHU-DVL/AttackIQA.

SPMGAE: Self-purified masked graph autoencoders release robust expression power

To tackle the scarcity of labeled graph data, graph self-supervised learning (SSL) has branched into two paradigms: Generative methods and Contrastive methods. Inspired by MAE and BERT in computer vision (CV) and natural language processing (NLP), masked graph autoencoders (MGAEs) are gaining popularity in the generative genre. However, prevailing MGAEs are mostly designed under the assumption that the data has high homophilic score and is out of adversarial distortion. When people deliberately improve the performance on homophilic graph datasets, they ignore a critical issue that both internal heterophily and artificial attack noise are quite common in the real world. Therefore, when data itself is highly heterophilic or confronted with attacks, they merely have no defensive capability. Especially under self-supervised conditions, it is much more difficult to detect internal heterophily and resist artificial attacks. In this paper, we propose a Self-Purified Masked Graph Autoencoder (SPMGAE) to make up for the shortcomings of prevailing MGAEs in terms of robustness. SPMGAE first utilizes a self-purified module to prune raw graph data and separate perturbation information. The purified graph provides a robust graph structure for the entire pre-training process. Next, the encoding module reuses perturbation information for auxiliary training to enhance robustness, while the decoding module reconstructs the effective graph data at a finer granularity. Extensive experiments on homophilic and heterophilic datasets attacked by various attack methods demonstrate SPMGAE has a considerable robust expressive ability. Especially on small datasets with large perturbations, the improvement of defensive performance could reaches 10%–25%.

MPSD: A Robust Defense Mechanism against Malicious PowerShell Scripts in Windows Systems

This manuscript introduces MPSD (Malicious PowerShell Script Detector), an advanced tool to protect Windows systems from malicious PowerShell commands and scripts commonly used in fileless malware attacks. These scripts are often hidden in Office document macros or downloaded remotely via PowerShell, posing significant threats to corporate networks. A 2018 report revealed that 77% of successful cyberattacks involved fileless malware, with PowerShell being the primary attack method, as highlighted in Red Canary’s 2022 report. To counter these threats, MPSD leverages the Antimalware Scan Interface (AMSI) to intercept and analyze real-time PowerShell scripts, preventing their execution. It further utilizes VirusTotal to filter out malicious scripts. Unlike traditional methods that rely on direct access to scripts, MPSD detects them before execution, addressing the challenge of hidden or obfuscated scripts. Experimental results show that MPSD outperforms well-known antivirus engines, with a low false-negative rate of 1.83%. MPSD is highly effective against evasion techniques like concatenation, encoding, and reordering, making it a robust tool in the cybersecurity landscape.

Adversarial attacks and defenses on text-to-image diffusion models: A survey

Recently, the text-to-image diffusion model has gained considerable attention from the community due to its exceptional image generation capability. A representative model, Stable Diffusion, amassed more than 10 million users within just two months of its release. This surge in popularity has facilitated studies on the robustness and safety of the model, leading to the proposal of various adversarial attack methods. Simultaneously, there has been a marked increase in research focused on defense methods to improve the robustness and safety of these models. In this survey, we provide a comprehensive review of the literature on adversarial attacks and defenses targeting text-to-image diffusion models. We begin with an overview of text-to-image diffusion models, followed by an introduction to a taxonomy of adversarial attacks and an in-depth review of existing attack methods. We then present a detailed analysis of current defense methods that improve model robustness and safety. Finally, we discuss ongoing challenges and explore promising future research directions. For a complete list of the adversarial attack and defense methods covered in this survey, please refer to our curated repository at https://github.com/datar001/Awesome-AD-on-T2IDM. Warning:This paper includes model-generated content that may contain offensive or distressing material.

Efficient Attacks on Strong PUFs via Covariance and Boolean Modeling

The physical unclonable function (PUF) is a widely used hardware security primitive. Before hacking into a PUF-protected system, intruders typically initiate attacks on the PUF as the first step. Many strong PUF designs have been proposed to thwart non-invasive attacks that exploit acquired CRPs. In this work, we propose a general framework for efficient attacks on strong PUFs by investigating from two perspectives, namely, statistical covariances in the challenge space and the design dependency among PUF compositions. The framework consists of two novel attack methods against a wide range of PUF families, including XOR APUFs, interpose PUFs, and bistable ring (BR)-PUFs. It can also exploit the knowledge of reliability information to improve attack efficiency with gradient optimization. We evaluate our proposed attacks through extensive experiments, running both software-based simulation and hardware implementations on FPGAs to compare with corresponding SOTA works. Considerable effort has been made in ensuring identical software/hardware conditions for a fair comparison. The results demonstrate that our framework significantly outperforms SOTA results. Moreover, we show that our framework can efficiently attack diverse PUF families built from entirely different types, while almost all existing works solely focused on attacking one or very limited number of PUF designs.

Digital Communication Crisis: The Case of Patiswiss Brand

Developments in information and communication technologies have intensely changed the way individuals communicate. The increase in technological innovations has led to the emergence of some risks for businesses, as well as the advantages it offers. Chief among these risks is that individuals have begun to easily express their expectations and complaints -especially in the context of the products they buy and use- through social media platforms where interaction and content sharing are possible. An individual who spends a significant amount of his time in the digital environment is both exposed to the content created by other users and can have an impact on many people by producing content. Under these conditions, social media environments and consumers/customers who use these environments intensively have positive effects on companies and brands at a level that will increase sales, and negative effects at a level that may cause a crisis. Managing technology-mediated communication environments is important in terms of managing such effects. The aim of this study is to reveal how the crisis is managed and/or attempted to be managed through digital media, within the framework of the need to manage the crisis that emerges through a content, issue, event that comes to the fore regarding the company(ies) in the digital media and spreads rapidly among the masses. In the study, which focused on digital communication, crisis management was discussed through a case study of the Patiswiss brand over the series of events that broke out in April 2024. The headlines and spots of the news published in the digital environment regarding the crisis process, developments and resolution of the crisis were analyzed through 5W1H. As a result, it has been observed that the communication language was transformed by developing a successful strategy after the attack and threat based communication method used in the beginning.

RPKI Defense Capability Simulation Method Based on Container Virtualization

As the main inter-domain routing protocol in today’s internet, the Border Gateway Protocol (BGP) faces serious security risks during actual usage. Research on BGP malicious attack methods requires a realistic network environment, and evaluation methods based on physical networks often suffer from high costs and insufficient flexibility. Thus, we propose an efficient BGP simulated network deployment system based on a virtualization technology called the SOD–BGP. This system, combining cloud computing and virtualization technologies, creates a scalable, highly flexible basic network environment that allows for the automated simulation and evaluation of actual BGP prefix hijacking attack scenarios. A Resource Public Key Infrastructure (RPKI) simulation suite is introduced into the system, emulating a certificate issuance system, certificate storage, and a certificate synchronization verification mechanism, thus aligning the simulation environment with real-world usage scenarios. Finally, we propose a data collection and performance evaluation technique to evaluate BGP networks deploying RPKI under different attack scenarios and to explore the effectiveness of RPKI defense mechanisms at various deployment rates. A comparative analysis with other simulation techniques demonstrates that our approach achieves a balanced performance in terms of deployment speed, complexity, and RPKI integrity, providing a solid simulation technology foundation for large-scale BGP security defense strategies.

Risk‐Aware SDN Defense Framework Against Anti‐Honeypot Attacks Using Safe Reinforcement Learning

ABSTRACTThe development of multiple attack methods by external attackers in recent years poses a huge challenge to the security and efficient operation of software‐defined networks (SDN), which are the core of operational controllers and data storage. Therefore, it is critical to ensure that the normal process of network interaction between SDN servers and users is protected from external attacks. In this paper, we propose a risk‐aware SDN defense framework based on safe reinforcement learning (SRL) to counter multiple attack actions. Specifically, the defender uses SRL to maximize the utility by choosing to provide a honeypot service or pseudo‐honeypot service within predefined security constraints, while the external attacker maximizes the utility by choosing an anti‐honeypot attack or masquerade attack. To describe the system risk in detail, we introduce the risk level function to model the simultaneous dynamic attack and defense processes. Simulation results demonstrate that our proposed risk‐aware scheme improves the defense utility by 17.5% and 142.4% and reduces the system risk by 42.7% and 59.6% compared to the QLearning scheme and the Random scheme, respectively.

Towards universal and transferable adversarial attacks against network traffic classification

In recent years, deep learning technology has shown astonishing potential in many fields, but at the same time, it also hides serious vulnerabilities. In the field of network traffic classification, attackers exploit this vulnerability to add designed perturbations to normal traffic, causing incorrect network traffic classification to implement adversarial attacks. The existing network traffic adversarial attack methods mainly target specific models or sample application scenarios, which have many problems such as poor transferability, high time cost, and low practicality. Therefore, this article proposes a method towards universal and transferable adversarial attacks against network traffic classification, which can not only perform universal adversarial attacks on all samples in the network traffic dataset, but also achieve cross data and cross model transferable adversarial attacks, that is, it has transferable attack effects at both the network traffic data and classification model levels. This method utilizes the geometric characteristics of the network model to design the target loss function and optimize the generation of universal perturbations, resulting in biased learning of features at each layer of the network model, leading to incorrect classification results. Meanwhile, this article conducted universality and transferability adversarial attack verification experiments on standard network traffic datasets of three different classification applications, USTC-TFC2016, ISCX2016, and CICIoT2023, as well as five common network models such as LeNet5. The results show that the proposed method performs universal adversarial attacks on five network models on three datasets, USTC-TFC2016, ISCX2016, and CICIoT2023, with an average attack success rate of over 80 %, 85 %, and 88 %, respectively, and an average time cost of about 0–0.3 ms; And the method proposed in this article has shown good transferable attack performance between five network models and on three network traffic datasets, with transferable attack rates approaching 100 % across different models and datasets, which is more closely related to practical applications.

SSBM: A spatially separated boxes-based multi-tab website fingerprinting model

In recent years, the website fingerprinting (WF) attack against the Tor anonymity system has become a hot research issue. The state-of-the-art WF studies have shown that the detection accuracy of websites is up to more than 95%. However, they are mainly conducted under the single-tab assumption, where each sample contains only one website traffic. The single-tab setting could not be realistic because users often open multiple tabs to browse simultaneously. The requests and responses from multiple tabs will overlap and interfere with each other, destroying existing single-tab WF attacks. In addition, the proposed multi-tab WF attack works poorly when traffic overlaps seriously. It remains challenging to implement WF attacks in multi-tab scenarios. This paper investigates a new spatial separated boxes-based multi-tab website fingerprinting model, called SSBM, to solve the multi-tab WF problem. It is an end-to-end model that separates traffic by equal-sized boxes and extracts features with convolutional neural networks. By predicting the label of each box, the tabs of the whole traffic are inferred. We design and implement SSBM and compare it with state-of-the-art multi-tab WF attacks in two different multi-tab modes: overlapping mode and delayed mode. In the overlapping mode, SSBM can successfully identify 81.24% of the first tab and 64.72% of the second tab when the overlapping proportions of the two tabs’ traffic reaches 50%, which are 4% and 29% higher than the current strongest BAPM. In the delayed mode, when the second tab traffic starts to overlap with the first tab traffic with a 5-second delay, SSBM improves the first tab’s classification accuracy from 60% to 69% and the second tab’s detection rates from 33% to 53%. Moreover, SSBM achieves the highest improvement, nearly 40%, in the three-tab evaluations. The experimental results show that SSBM outperforms existing multi-tab WF attack methods.

Attack-resilient framework for wind power forecasting against civil and adversarial attacks

Forecasting wind power generation accurately is crucial for reliable, economical, and efficient integrations in smart grids, promoting applications of cleaner energy sources. Although effective wind power forecasting methods exist, power grids still require resilient schemes enabling accurate predictions under cyber-attacks. This paper introduces civil attack (CA) and fast gradient sign method (FGSM) attacks to wind power forecasting to analyze their impacts with countermeasures. The impacts of CA and FGSM attacks on a deep learning-based forecasting method are evaluated, finding FGSM attacks more severe. Also, an attack identification and corrupted data replacement-based pre-processing robust framework is proposed, outperforming other countermeasures. To detect and classify attacks, random forest (RF) has outperformed extreme gradient boosting (XGBoost), decision tree (DT), support vector machine (SVM), and k-nearest neighbors (KNN). Experimental results on two different zones during CA and FGSM attacks indicate that the decrease in accuracy can be up to 0.4103, 0.3152, and 0.1683 in terms of root mean square error (RMSE), mean absolute error (MAE), and mean squared error (MSE), respectively. The proposed framework successfully achieves an accuracy of 0.1204, 0.0835, and 0.0145 for the worst case in terms of RMSE, MAE, and MSE, respectively, signifying its importance for academic and industrial applications.

Mixed training backdoor attack method improves generalization of backdoor attacks: An empirical study based on multiple models and datasets

In recent years, backdoor attacks have posed significant security threats to the training process of deep neural networks (DNNs). Attackers attempt to embed triggers in the training set, causing the victim model to behave normally when processing benign samples, but maliciously alter predictions when the hidden backdoor is activated by these triggers. However, our research found that when the trigger locations differ between the training and test data, the success rate of backdoor attacks decreases. This finding was consistent across different models and datasets through our experiments. To address the issue of decreased attack success rates caused by inconsistent trigger distributions between training and test data, we propose a mixed training backdoor attack method. This method enhances the generalization of attacks by constructing backdoor attack data with triggers in various positions during the training phase. When applying this mixed training backdoor attack approach to the BadNets and Blended attack methods, the ASR (Attack Success Rate) on test sets improved by up to 99.57% and 94.96%, respectively, across different models and datasets.

End-to-End Learning-Based Study on the Mamba-ECANet Model for Data Security Intrusion Detection

With the rapid development of information technology, network security issues have become increasingly prominent. In particular, data security intrusions pose serious threats to the data privacy and system security of enterprises and individuals. Traditional intrusion detection systems often exhibit low detection accuracy and high false alarm rates when faced with complex and dynamic network environments and diverse attack methods. Therefore, this paper proposes a data security intrusion detection system based on deep learning, which integrates the Mamba model and ECANet model and employs an end-to-end learning approach for training and optimization. First, the Mamba model is introduced for preliminary data feature extraction, whose efficient feature representation capabilities provide a solid foundation for the subsequent detection process. Then, by integrating the ECANet model, feature selection is further optimized through the attention mechanism, enhancing the model’s focus on important features. Finally, an end-to-end learning approach is adopted to train and optimize the entire system, ensuring excellent performance and robustness in practical applications. Experimental results show that the proposed intrusion detection system demonstrates higher detection accuracy on multiple test datasets, improving by approximately 5% compared to traditional methods, providing a new and effective solution for data security.