Diverse Attacks Research Articles

Adversarial Attacks in Explainable Machine Learning: A Survey of Threats Against Models and Humans

ABSTRACTReliable deployment of machine learning models such as neural networks continues to be challenging due to several limitations. Some of the main shortcomings are the lack of interpretability and the lack of robustness against adversarial examples or out‐of‐distribution inputs. In this paper, we comprehensively review the possibilities and limits of adversarial attacks for explainable machine learning models. First, we extend the notion of adversarial examples to fit in explainable machine learning scenarios where a human assesses not only the input and the output classification, but also the explanation of the model's decision. Next, we propose a comprehensive framework to study whether (and how) adversarial examples can be generated for explainable models under human assessment. Based on this framework, we provide a structured review of the diverse attack paradigms existing in this domain, identify current gaps and future research directions, and illustrate the main attack paradigms discussed. Furthermore, our framework considers a wide range of relevant yet often ignored factors such as the type of problem, the user expertise or the objective of the explanations, in order to identify the attack strategies that should be adopted in each scenario to successfully deceive the model (and the human). The intention of these contributions is to serve as a basis for a more rigorous and realistic study of adversarial examples in the field of explainable machine learning.

Domain knowledge free cloud-IDS with lightweight embedding method

The expansion of the cloud computing market has provided a breakthrough in efficiently storing and managing data for individuals and companies. As personal and corporate data move to the cloud, diverse attacks targeting the cloud have also increased for heist beneficial information. Therefore, cloud service providers offer protective environments through diverse security solutions. However, security solutions are limited in preventing advanced attacks because it is challenging to reflect the environment of each user. This paper proposes a Cloud Intrusion Detection System (C-IDS) that adapts to each user’s cloud environment and performs real-time attack detection using Natural Language Processing (NLP). Notably, the C-IDS learns the deployed client environment logs and detects anomalies using the Seq2Seq model with BI-LSTM and Bahdanau attention. We used multiple domain datasets, Linux, Windows, Hadoop, OpenStack, Apache, OpenSSH, and CICIDS2018 to verify the performance of the C-IDS. C-IDS consists of a ‘recognition’ that identifies logs in the deployed environment and a ‘detection’ that discovers anomalies. The recognition results showed an average accuracy of 98.2% for multiple domain datasets. Moreover, the detection results based on the trained model exhibited an average accuracy of 94.2% for the Hadoop, OpenStack, Apache, and CICIDS2018 datasets.

A Blockchain-Based Security Framework for East-West Interface of SDN

Software-Defined Networking (SDN) has emerged as a revolutionary architecture in computer networks, offering comprehensive network control and monitoring capabilities. However, securing the east–west interface, which is crucial for communication between distributed SDN controllers, remains a significant challenge. This study proposes a novel blockchain-based security framework that integrates Ethereum technology with customized blockchain algorithms for authentication, encryption, and access control. The framework introduces decentralized mechanisms to protect against diverse attacks, including false data injection, man-in-the-middle (MitM), and unauthorized access. Experimental results demonstrate the effectiveness of this framework in securing distributed controllers while maintaining high network performance and low latency, paving the way for more resilient and trustworthy SDN infrastructures.

KRF-AD: Innovating anomaly detection with KDE-KL and random forest fusion

Anomaly detection in Intrusion Detection System (IDS) data refers to the process of identifying and flagging unusual or abnormal behavior within a network or system. In the context of IoT, anomaly detection helps in identifying any abnormal or unexpected behavior in the data generated by connected devices. Existing methods often struggle with accurately detecting anomalies amidst massive data volumes and diverse attack patterns. This paper proposes a novel approach, KDE-KL Anomaly Detection with Random Forest Integration (KRF-AD), which combines Kernel Density Estimation (KDE) and Kullback-Leibler (KL) divergence with Random Forest (RF) for effective anomaly detection. Additionally, Random Forest (RF) integration enables classification of data points as anomalies or normal based on features and anomaly scores. The combination of statistical divergence measurement and density estimation enhances the detection accuracy and robustness, contributing to more effective network security. Experimental results demonstrate that KRF-AD achieves 96% accuracy and outperforms other machine learning models in detecting anomalies, offering significant potential for enhancing network security.

An effective attack detection framework using multi-scale depth-wise separable 1DCNN via fused grasshopper-based lemur optimizer in IoT routing system

A secured IoT routing model against different attacks has been implemented to detect attacks like replay attacks, version attacks, and rank attacks. These attacks cause certain issues like energy depletion, minimized packet delivery, and loop creation. By mitigating these issues, an advanced attack detection approach for secured IoT routing techniques with a deep structured scheme is promoted to attain an efficient attack detection rate over the routing network. In the starting stage, the aggregation of data is done with the help of IoT networks. Then, the selected weighted features are subjected to the Multiscale Depthwise Separable 1-Dimensional Convolutional Neural Networks (MDS-1DCNN) approach for attack detection, in which the parameters in the 1-DCNN are tuned with the aid of Fused Grasshopper-aided Lemur Optimization Algorithm (FG-LOA). The parameter optimization of the FG-LOA algorithm is used to enlarge the efficacy of the approach. Especially, the MDS-1DCNN model is used to detect different attacks in the detection phase. The attack nodes are mitigated during the routing process using the developed FG-LOA by formulating the fitness function based on certain variables such as shortest distance, energy, path loss and delay, and so on in the routing process. Finally, the performances are examined through the comparison with different traditional methods. From the validation of outcomes, the accuracy value of the developed attack detection model is 96.87%, which seems to be better than other comparative techniques. Also, the delay analysis of the routing model based on FG-LOA is 17.3%, 12.24%, 10.41%, and 15.68% more efficient than the classical techniques like DHOA, HBA, GOA, and LOA, respectively. Hence, the effectualness of the offered approach is more enriched than the baseline approaches and also it has mitigated diverse attacks using secured IoT routing and different attack models.

Blockchain-Machine Learning Fusion for Enhanced Malicious Node Detection in Wireless Sensor Networks

In wireless sensor networks (WSNs), the presence of malicious nodes (MNs) poses significant challenges to data integrity, network stability, and system reliability. These issues are intensified by energy resource constraints and limitations within centralized authentication systems, necessitating an energy-efficient solution to ensure real-time responsiveness. Although artificial intelligence-driven approaches enhance detection capabilities, they overcome challenges related to data volume, coordination overhead, and latency in centralized control. This study introduces blockchain-machine learning (BC-ML), a novel hybrid model that seamlessly integrates blockchain and machine learning (ML) techniques to effectively identify MNs in WSNs. The model establishes an energy-efficient blockchain among cluster heads (CHs) for robust node authentication, incorporating a Schnorr-like zero-knowledge-proof technique to validate node data during communication initiation. Utilizing a hybrid lightweight approach with both symmetric and asymmetric ciphers enhances the security of node data transmission. A new proof-of-authority method is introduced, which leverages node digital certificates instead of conventional data transactions. This consensus mechanism reduces the processing overhead associated with larger data sizes in traditional proof-of-work methods, thereby improving both energy efficiency and scalability. To address dataset imbalances, the model employs a hybrid unsupervised ML technique, combining adaptive synthetic sampling with a convolutional neural network for efficient analysis of nodes and network features. The ML model, hosted on a robust data server, ensures ongoing oversight by updating CHs with security levels for detected MNs, thereby reducing storage and mitigating coordination challenges. Comprehensive analyses validate the effectiveness of the BC-ML model for detecting MNs, optimizing resource utilization, minimizing delays, and prolonging node and network lifetimes. Security analysis further confirms the ability of the model to mitigate diverse attacks and meet the stringent WSN security requirement.

Detection of Intrusions using Transformer - LSTM Model

The project focuses on addressing the escalating issue of web attacks and ensuring cyberspace security in the era of advancing internet technology. The primary goal is to enhance intrusion detection systems (IDS) by moving beyond traditional keyword and rule- based methods. The project explores the integration of artificial intelligence, specifically deep learning techniques, to develop a more robust and adaptive approach to identifying network intrusion events. With the limitations of conventional detection methods in mind, there is a critical need for a more effective and versatile intrusion detection system. The rise in diverse attack methods necessitates an intelligent system capable of autonomously adapting to new attack patterns without relying on manual rule creation. This project is designed to benefit both individuals and organizations reliant on network services. Improved intrusion detection ensures a higher level of cybersecurity, safeguarding sensitive data and systems. The project's outcomes have the potential to contribute significantly to the broader field of network security. Leveraging the power of deep learning, the proposed model- Transformer (autoencoder) LSTM combines Transformer and LSTM architectures. Multiple attention mechanisms are employed to select and extract features, while the LSTM model facilitates understanding the sequential relationships within network traffic, enabling the accurate identification of intrusion events. And we are comparing the model with DNN, LSTM and Transformer(autoencoder). To enhance performance, CNN and CNN+LSTM extensions were incorporated into the project, complementing the Transformer (autoencoder) LSTM model. These models aim to further refine feature extraction and exploit spatial relationships within network traffic. Key Words:intrusion detection, transformer, LSTM, deep learning.

A Practical Approach to Intrusion Detection in IoV Networks Using LCCDE Ensemble Framework

Abstract: With the rapid development of the Internet of Things (IoT) and the Internet of Vehicles (IoV) technologies, modern vehicles are increasingly adopting network-controlled functionalities, exposing them to a variety of cyber threats. To address these security challenges, this paper implements and evaluates a novel ensemble Intrusion Detection System (IDS) framework, named Leader Class and Confidence Decision Ensemble (LCCDE). The LCCDE framework integrates three advanced Machine Learning (ML) algorithms—XGBoost, LightGBM, and CatBoost—to detect various types of cyber-attacks on both intra-vehicle and external vehicular networks.This study demonstrates the effectiveness of the LCCDE framework using two public IoV security datasets: the Car-Hacking dataset and the CICIDS2017 dataset. By identifying the best-performing ML model for each class of attacks and leveraging prediction confidence values, LCCDE achieves high detection accuracy and robustness against diverse attack types. The experimental results show that the proposed framework outperforms traditional IDS approaches in terms of detection accuracy, computational efficiency, and adaptability to different types of cyber-attacks. This paper provides practical insights into the implementation and deployment of IDS in IoV systems, highlighting the potential of ensemble learning methods to enhance vehicular network security.

Enhancing IoT Routing Security and Efficiency: Towards AI-Enabled RPL Protocol

In the rapidly evolving landscape of the Internet of Things (IoT), the security and efficiency of network routing are paramount. The standard Routing Protocol for Low-Power and Lossy Networks (RPL) faces challenges in the form of various cyber-attacks, impacting its reliability and performance. To address this problem, we propose a Novel Algorithm for Network Traffic Analysis and Response (NANTAR), a novel algorithm that optimizes the security and efficiency of RPL routing using AI-based techniques. Through evaluation against traditional RPL, NANTAR has demonstrated remarkable improvements in key performance metrics: a 20% increase in throughput, latency reductions of 20-30%, and more efficient resource utilization. Notably, NANTAR maintains robust network performance with scaling, evidenced by 15-20% lower false positive and negative rates. The algorithm’s efficacy is further highlighted by its high detection rates in the face of diverse attacks and its power consumption efficiency, which is crucial for IoT devices. NANTAR’s adaptability and seamless integration across various IoT platforms affirm its potential as a solution for securing IoT ecosystems, with results indicating significant improvements in key operational metrics.

Developing a secure image encryption technique using a novel S-box constructed through real-coded genetic algorithm’s crossover and mutation operators

The objective of this study is to craft a novel S-Box tailored to stringent security standards, achieved through iterative application of crossover and mutation operators inherent to real-coded genetic algorithms, ensuring robust image encryption. The designed S-Box is rigorously evaluated across multiple criteria, demonstrating its superiority over comparable S-Box designs found in existing literature. Furthermore, a secure image encryption method based on the designed S-Boxes devised and also including a 2D hyperchaotic Styblinski–Tang map. Thorough security assessments, encompassing statistical analysis and resilience testing against diverse attacks and noise, it is validated the encryption technique’s efficiency and applicability across varied scenarios.

IDS-DEC: A novel intrusion detection for CAN bus traffic based on deep embedded clustering

As the automotive industry advances towards greater automation, the proliferation of electronic control units (ECUs) has led to a substantial increase in the connectivity of in-vehicle networks with the external environment. However, the widely used Controller Area Network (CAN), which serves as the standard for in-vehicle networks, lacks robust security features, such as authentication or encrypted information transmission. This poses a significant challenge to the security of these networks. Despite the availability of powerful intrusion detection methods based on machine learning and deep learning, there are notable limitations in terms of stability and accuracy in the absence of a supervised learning process with labeled data. To address this issue, this paper introduces a novel in-vehicle intrusion detection system, termed IDS-DEC. This system combines a spatiotemporal self-coder employing LSTM and CNN (LCAE) with an entropy-based deep embedding clustering. Specifically, our approach involves encoding in-vehicle network traffic into windowed messages using a stream builder, designed to adapt to high-frequency traffic. These messages are then fed into the LCAE to extract a low-dimensional nonlinear spatiotemporal mapping from the initially high-dimensional data. The resulting low-dimensional mapping is subjected to a dual constraint in conjunction with our entropy-based pure deep embedding clustering module. This creates a bidirectional learning objective, addressing the optimization problem and facilitating an end-to-end training pattern for our model to adapt to diverse attack environments. The effectiveness of IDS-DEC is validated using both the benchmark Car Hacking dataset and the Car Hacking-Attack & Defense Challenge dataset. Experimental results demonstrate the model's high detection accuracy across various attacks, stabilizing at approximately 99% accuracy with a 0.5% false alarm rate. The F1 score also stabilizes at around 99%. In comparison with unsupervised methods based on deep stream clustering, LSTM-based self-encoder, and classification-based methods, IDS-DEC exhibits significant improvements across all performance metrics.

Perturbation Augmentation for Adversarial Training with Diverse Attacks

Adversarial Training (AT) aims to alleviate the vulnerability of deep neural networks to adversarial perturbations. However, the AT techniques struggle to maintain the performance on natural samples while improving the deep model’s robustness. The absence of perturbation diversity in generated during the adversarial training degrades the generalizability of the robust models, causing overfitting to particular perturbations and a decrease in natural performance. This study proposes an adversarial training framework that augments adversarial directions from a single-step attack to address the trade-off between robustness and generalization. Inspired by feature scattering adversarial training, the proposed framework computes a principal adversarial direction with a single-step attack that finds a perturbation disrupting the inter-sample relationships in the mini-batch during adversarial training. The principal direction obtained at each iteration is augmented by sampling new adversarial directions within a region spanning 45 degrees from the principal adversarial direction. The proposed adversarial training approach does not require extra backpropagation steps in adversarial direction augmentation. Therefore, generalization of the robust model is improved without posing an additional burden on the feature scattering adversarial training. Experiments on CIFAR-10, CIFAR-100, SVHN, Tiny-ImageNet, and The German Traffic Sign Recognition Benchmark consistently improve the accuracy on adversarial with an almost pristine natural performance.

An innovative multi-agent approach for robust cyber–physical systems using vertical federated learning

Federated learning presents a compelling approach to training artificial intelligence systems in decentralized settings, prioritizing data safety over traditional centralized training methods. Understanding correlations among higher-level threats exhibiting abnormal behavior in the data stream becomes paramount to developing cyber–physical systems resilient to diverse attacks within a continuous data exchange framework. This work introduces a novel vertical federated multi-agent learning framework to address the challenges of modeling attacker and defender agents in stationary and non-stationary vertical federated learning environments. Our approach uniquely applies synchronous Deep Q-Network (DQN) based agents in stationary environments, facilitating convergence towards optimal strategies. Conversely, in non-stationary contexts, we employ synchronous Advantage Actor–Critic (A2C) based agents, adapting to the dynamic nature of multi-agent vertical federated reinforcement learning (VFRL) environments. This methodology enables us to simulate and analyze the adversarial interplay between attacker and defender agents, ensuring robust policy development. Our exhaustive analysis demonstrates the effectiveness of our approach, showcasing its capability to learn optimal policies in both static and dynamic setups, thus significantly advancing the field of cyber-security in federated learning contexts. To evaluate the effectiveness of our approach, we have done a comparative analysis with its baseline schemes. The findings of our study show significant enhancements compared to the standard methods, confirming the efficacy of our methodology. This progress dramatically enhances the area of cyber-security in the context of federated learning by facilitating the formulation of substantial policies. The proposed scheme attains 15.93%, 32.91%, 31.02%, and 47.26% higher results as compared to the A3C, DDQN, DQN, and Reinforce, respectively.

Resiliency of forecasting methods in different application areas of smart grids: A review and future prospects

The cyber–physical infrastructure of a smart grid requires data-dependent artificial intelligence (AI)-based forecasting schemes for predicting different aspects for the short- to long-term, where AI-based schemes include machine learning (ML), deep learning (DL), and hybrid models. These forecasting schemes in different application areas of a smart grid can be vulnerable to cyber-attacks, which is yet to be addressed from a broad perspective. This work reviews the literature addressing the vulnerability of forecasting schemes in smart grids with a categorization of application areas. The existing research works addressing cyber-security or cyber resiliency are reviewed and then presented in an organized manner according to application areas to highlight their advantages and disadvantages. The findings of this review indicate a critical need to develop accurate and robust AI-based forecasting schemes capable of withstanding diverse attack scenarios in each sector, while addressing unsymmetrical attention to different sectors of smart grids. Hence, this review provides a comprehensive overview of the current literature and emphasizes the necessity for the research community to advance toward developing attack-resilient AI-based forecasting schemes designed explicitly for smart grids.

GANFAT: Robust federated adversarial learning with label distribution skew

As privacy concerns and regulatory constraints on data protection continue to grow, the distribution of collected data has become more dispersed, resembling a ”data silo” style. To harness these data effectively without exchanging raw data, federated learning has emerged as a prominent solution. However, distributions of user-generated data often exhibit imbalances between devices and labels, which adversely affect model performance, especially in the presence of adversarial attacks, making models more susceptible. To address the challenge of balancing natural accuracy and robustness in federated training, especially under skewed label distribution scenarios, we propose a novel approach based on Generative Adversarial Networks for Federated Adversarial Training (GANFAT). GANFAT leverages GAN to enhance the authenticity and effectiveness of adversarial samples and addresses label distribution skew issues by incorporating class probability distribution information. Through a balanced interplay of natural accuracy loss and adversarial loss, GANFAT demonstrates significantly superior performance across multiple datasets under various settings compared to other frameworks. Particularly on the SVHN dataset, GANFAT achieves a remarkable 9.30% enhancement in robustness against FGSM attacks compared to the best baseline method (FedRBN). On the CIFAR-100 dataset, GANFAT showcases a noteworthy 6.68% improvement in natural accuracy compared to the best baseline method (CalFAT). GANFAT provides a powerful solution for confronting diverse attacks, yielding models comparable to those produced by centralized training. Experimental results underscore GANFAT’s outstanding performance, offering a robust solution for scenarios characterized by uneven data distribution and adversarial attacks.

Enhancing IoT security: A comparative study of feature reduction techniques for intrusion detection system

Internet of Things (IoT) devices are extensively utilized but are susceptible to cyberattacks, posing significant security challenges. To mitigate these threats, machine learning techniques have been implemented for network intrusion detection in IoT environments. These techniques commonly employ various feature reduction methods, prior to inputting data into models, in order to enhance the efficiency of detection processes to meet real-time requirements. This study provides a comprehensive comparison of feature selection (FS) and feature extraction (FE) techniques for network intrusion detection systems (NIDS) in IoT environments, utilizing the TON-IoT and BoT-IoT datasets for both binary and multi-class classification tasks. We evaluated FS methods, including Pearson correlation and Chi-square, and FE methods, such as Principal Component Analysis (PCA) and Autoencoders (AE), across five classic machine learning models: Decision Tree (DT), Random Forest (RF), Naive Bayes (NB), k-Nearest Neighbors (kNN), and Multi-Layer Perceptron (MLP). Our analysis revealed that FE techniques generally achieve higher accuracy and robustness compared to FS methods, with RF paired with AE delivering superior performance despite higher computational demands. DTs are most effective with smaller feature sets, while MLPs excel with larger sets. Chi-square is identified as the most efficient FS method, balancing performance and computational efficiency, whereas PCA outperforms AE in runtime efficiency. The study also highlights that FE methods are more effective for complex datasets and less sensitive to feature set size, whereas FS methods show significant performance improvements with more informative features. Despite the higher computational costs of FE methods, they demonstrate a greater capability to detect diverse attack types, making them particularly suitable for complex IoT environments. These findings are crucial for both academic research and industry applications, providing insights into optimizing detection performance and computational efficiency in NIDS for IoT networks.

A Systematic Deconstruction of Human-Centric Privacy & Security Threats on Mobile Phones

Mobile phones are most likely the subject of targeted attacks, such as software exploits. The resources needed to carry out such attacks are becoming increasingly available and, hence, easily executable, putting users’ privacy at risk. We conducted a systematic literature analysis to understand the relationship between resources and attack feasibility and present a categorisation of social engineering and side-channel attacks on mobile phones focusing on the resources attackers require. Our proposed categorisation levels facilitate an in-depth understanding of how mobile phone attacks can be executed using different combinations of partly simple resources. The analysis reveals that discrete protection mechanisms are insufficient to provide all-inclusive protection. The proposed categorisation assists in building novel solutions for safeguarding users’ privacy from diverse attacks by carefully considering the potential misuse of resources. We conclude by outlining future research directions highlighting the urgent need for a holistic user defense.

Cross-layer detection and defence mechanism against DDoS and DRDoS attacks in software-defined networks using P4 switches

This paper presents a comprehensive system for detecting and defending against distributed denial-of-service (DDoS) and distributed reflective denial-of-service (DRDoS) flood attacks in software-defined networking (SDN) environments using Programming Protocol-Independent Packet Processors (P4). The proposed system introduces a hybrid intrusion statistical threshold algorithm (HISTA) that analyzes intrusion data statistics to identify potential malicious attacks. Upon detection, the system utilises P4 programmability to implement a custom defence mechanism on the P4 switch. A novel Uruk protocol header collaborates with HISTA for enhanced cross-layer attack detection and categorisation. The HISTA-Uruk mechanism selectively discards malicious packets while ensuring uninterrupted communication for legitimate packets, effectively preventing DDoS/DRDoS attacks. Extensive experiments validate the system’s ability to efficiently detect and thwart various DDoS, DRDoS, and internal attacks, demonstrating its effectiveness in identifying threats and restoring impacted network connections across diverse attack scenarios.

DTT: A Dual-domain Transformer model for Network Intrusion Detection

With the rapid evolution of network technologies, network attacks have become increasingly intricate and threatening. The escalating frequency of network intrusions has exerted a profound influence on both industrial settings and everyday activities. This underscores the urgent necessity for robust methods to detect malicious network traffic. While intrusion detection techniques employing Temporal Convolutional Networks (TCN) and Transformer architectures have exhibited commendable classification efficacy, most are confined to the temporal domain. These methods frequently fall short of encompassing the entirety of the frequency spectrum inherent in network data, thereby resulting in information loss. To mitigate this constraint, we present DTT, a novel dual-domain intrusion detection model that amalgamates TCN and Transformer architectures. DTT adeptly captures both high-frequency and low-frequency information, thereby facilitating the simultaneous extraction of local and global features. Specifically, we introduce a dual-domain feature extraction (DFE) block within the model. This block effectively extracts global frequency information and local temporal features through distinct branches, ensuring a comprehensive representation of the data. Moreover, we introduce an input encoding mechanism to transform the input into a format suitable for model training. Experiments conducted on two distinct datasets address concerns regarding data duplication and diverse attack types, respectively. Comparative experiments with recent intrusion detection models unequivocally demonstrate the superior performance of the proposed DTT model.

Stealthy attacks against distributed state estimation of stochastic multi-agent systems under composite attack detection mechanisms

This paper is concerned with the security problem of distributed state estimation for stochastic multi-agent systems (MASs) under directed topology. First, in the absence of attacks, an optimal distributed filter is proposed to reconstruct the states of MASs by minimizing the mean square error, which makes full use of the measurable local output information and relative state estimates of agents. Moreover, instead of using a single detection mechanism to measure the stealthiness of attacks, based on the structure of the derived filter, a composite attack detection mechanism capable of simultaneously detecting anomalies in the sensor channels of the agents themselves and the network topologies between different agents is utilized in this paper. Then, from the perspective of potential adversaries, two stealthy attack strategies are established through a joint design of sensor and topology attacks, enabling attackers to fulfill diverse attack objectives while evading the given composite detection mechanisms. Finally, simulation results are provided to verify the availability of the designed filter and attack generation schemes.