Defense Model Research Articles

Impact of AI on the Cyber Kill Chain: A Systematic Review

The Cyber Kill Chain (CKC) defense model aims to assist subject matter experts in planning, identifying, and executing against cyber intrusion activity, by outlining seven stages required for adversaries to execute an attack. Recent advancements in Artificial Intelligence (AI) have empowered adversaries to execute sophisticated attacks to exploit system vulnerabilities. As a result, it is essential to consider how AI-based tools change the cyber threat landscape and affects the current standard CKC model. Thus, this study examines and categorizes how attackers use AI-based tools and offers potential defense mechanisms. We conducted a systematic literature review of 62 papers published between 2013 and 2023 from the Web of Science and Google Scholar databases. Our findings indicate that AI-based tools are used most effectively in the initial stages of cyberattacks. However, we find that current defence tools are not designed to counter these sophisticated attacks during these stages. Thus, we provide insights to 1) highlight the changing threat landscape due to AI and 2) to guide the development of cyberdefense mechanisms.

Exploring Synergy of Denoising and Distillation: Novel Method for Efficient Adversarial Defense

Escalating advancements in artificial intelligence (AI) has prompted significant security concerns, especially with its increasing commercialization. This necessitates research on safety measures to securely utilize AI models. Existing AI models are vulnerable to adversarial attacks, which are a specific form of assault methodology. Although various countermeasures have been explored, practical defense models are scarce. Current adversarial defense methods suffer from reduced accuracy, increased training time, and incomplete defense against adversarial attacks, indicating performance limitations and a lack of robustness. To address these limitations, we propose a composite defense model, the knowledge Distillation and deNoising Network (DiNo-Net), which integrates knowledge distillation and feature denoising techniques. Furthermore, we analyzed a correlation between the loss surface of adversarial perturbations and denoising techniques. Using DiNo-Net, we confirmed that increasing the temperature during the knowledge distillation process effectively amplifies the loss surface around the ground truth. Consequently, this enables more efficient denoising of the adversarial perturbations. It achieved a defense success rate of 72.7%, which is a remarkable improvement over the 41.0% success rate of models with only denoising defense mechanisms. Furthermore, DiNo-Net reduced the training time and maintained higher accuracy, confirming its efficient defense performance. We hope that this relationship will spur the development of fundamental defense strategies.

Cracking the code: Linking good modeling and coding practices for new ecological modelers

Good modeling practices are essential for producing reliable and reproducible ecological models. Inherent to good modeling practices are fundamental coding and documentation skills, which not only implement the desired modeling capabilities but also clearly outline the goals, methods, and components of a model that are necessary to reproduce the desired results. One of the largest challenges for new ecological modelers can be implementing a model into computer code. In fact, coding represents a significant barrier for entry into ecological modeling, since most ecologists have not had formal training in computer science or software development. While software packages do exist that facilitate model development, we have observed that newer modelers still struggle with developing good coding practice throughout the modeling process. During a series of agent-based modeling short-courses and full semester graduate courses, both taught in NetLogo, we identified some common challenges encountered by graduate students and environmental professionals as they learn to code an ecological model, many for the first time. We were able to categorize and provide examples of the main challenges and obstacles, which fell into three main groups that follow the steps of good modeling practice: problem scoping and conceptualization, formulation, and evaluation. We then provide guidance on how to overcome these obstacles while developing good coding and modeling practices that will result in more scientifically defensible models.

Defense against Adversarial Attacks in Image Recognition Based on Multilayer Filters

The security and privacy of a system are urgent issues in achieving secure and efficient learning-based systems. Recent studies have shown that these systems are susceptible to subtle adversarial perturbations applied to inputs. Although these perturbations are difficult for humans to detect, they can easily mislead deep learning classifiers. Noise injection, as a defense mechanism, can offer a provable defense against adversarial attacks by reducing sensitivity to subtle input changes. However, these methods face issues of computational complexity and limited adaptability. We propose a multilayer filter defense model, drawing inspiration from filter-based image denoising techniques. This model inserts a filtering layer after the input layer and before the convolutional layer, and incorporates noise injection techniques during the training process. This model substantially enhances the resilience of image classification systems to adversarial attacks. We also investigated the impact of various filter combinations, filter area sizes, standard deviations, and filter layers on the effectiveness of defense. The experimental results indicate that, across the MNIST, CIFAR10, and CIFAR100 datasets, the multilayer filter defense model achieves the highest average accuracy when employing a double-layer Gaussian filter (filter area size of 3×3, standard deviation of 1). We compared our method with two filter-based defense models, and the experimental results demonstrated that our method attained an average accuracy of 71.9%, effectively enhancing the robustness of the image recognition classifier against adversarial attacks. This method not only performs well on small-scale datasets but also exhibits robustness on large-scale datasets (miniImageNet) and modern models (EfficientNet and WideResNet).

Proactively Defending Enterprise Computer Systems Against Threats and Vulnerabilities

Cybersecurity remains a critical concern, even amidst global events like the COVID-19 pandemic. The rise of COVID-19 has deepened cybersecurity threats, with phishing emails and phone scams attempting to exploit the situation. This paper focuses on proactive strategies to protect enterprise environments against threats and vulnerabilities, specifically in Windows-based systems. Tracing threats and vulnerabilities to their source aims to address them in their early stages rather than after an attack has occurred. This research tackles three prevalent issues: phishing emails, vulnerability patching, and industrial internet-connected devices. Through analyzing various cyber defense models and vulnerability databases, this paper proposes frameworks to mitigate these issues effectively. The study includes a detailed examination of sources of threats and vulnerabilities, aiming to develop methodologies for practical implementation. Ultimately, the goal is to summarize best practices to enhance tool utilization and process improvement and propose new proactive defense methods. The research emphasizes the shift from reactive to proactive defense strategies to better protect enterprise networks.

Reducing non-technical losses in electricity distribution networks: Leveraging explainable AI and three lines of defence model to manage operational staff-related factors

This study presents a multidisciplinary approach involving Explainable Artificial Intelligence (ExAI) and operational risk management to reduce Non-Technical Losses (NTL) in electricity distribution. It empirically explores how the activities of employees of utility companies contribute to NTL, a phenomenon often overlooked in existing empirical research. An ensemble classification algorithm is used to analyse utility operations data, and the SHAP explainability technique establishes the predictive significance of staff activities for NTL. Subsequently, these staff activities are mapped into risk cells using the BASEL II and III operational risk definitions, and the Three Lines of Defence (3LoD) model is developed for optimizing electricity distribution. The paper makes three original contributions to the literature: first, it empirically links staff operations to NTL; second, it maps NTL causes to Basel II/III operational risk categories; and finally, to the best of the authors’ knowledge, it is the first study to use the 3LoD model for electricity distribution optimization.

Exploring Machine Learning Algorithms for Cyber Attacks’ Classification

Cyber security is like a race between defensive and offensive capabilities. Day by day the attacks are increasing consequently the preventive mechanisms are also raising. The technological evolution must address the cyber security problem effectively. Cyber security is always the cutting-edge technology to which Machine learning added potentiality. Machine learning plays a crucial role cyber-attacks in various dimensions. The recent advancements in usage of Machine learning towards cyber security preventive and detective measures clearly indicates that the Cyber security certainly accelerates. But at the same time the threat framework also takes the advent of innovative machine learning models. In this paper we are intended to explore various machine learning algorithms to classify cyber-attacks, which are very useful in designing efficient threat defensive models

Local carbon reserves are insufficient for phloem terpene induction during drought in Pinus edulis in response to bark beetle-associated fungi.

Stomatal closure during drought inhibits carbon uptake and may reduce a tree's defensive capacity. Limited carbon availability during drought may increase a tree's mortality risk, particularly if drought constrains trees' capacity to rapidly produce defenses during biotic attack. We parameterized a new model of conifer defense using physiological data on carbon reserves and chemical defenses before and after a simulated bark beetle attack in mature Pinus edulis under experimental drought. Attack was simulated using inoculations with a consistent bluestain fungus (Ophiostoma sp.) of Ips confusus, the main bark beetle colonizing this tree, to induce a defensive response. Trees with more carbon reserves produced more defenses but measured phloem carbon reserves only accounted for c. 23% of the induced defensive response. Our model predicted universal mortality if local reserves alone supported defense production, suggesting substantial remobilization and transport of stored resin or carbon reserves to the inoculation site. Our results show that de novo terpene synthesis represents only a fraction of the total measured phloem terpenes in P. edulis following fungal inoculation. Without direct attribution of phloem terpene concentrations to available carbon, many studies may be overestimating the scale and importance of de novo terpene synthesis in a tree's induced defense response.

NATIONAL SECURITY AND THE TOOLS OF CYBER POWER: A REVIEW ON THE AREAS OF STATE HEGEMONIA

In the post-Cold War era, in terms of international relations, 'cyber security' emerges as a significant security issue for the spheres of state sovereignty beyond the personal sphere. In addition to the harmful effects of cyber threats on the functioning of public institutions and organizations, İt also raises a problem for states, such as managing perception operations by creating a society more open to manipulative influences. In our globalizing world, as the most important actor of the system, the States' dependence on each other has increased, and at the same time, there has been a period in which the fundamental security issues are intertwined for the States. In such a situation where it is not possible for the States to define themselves outside the system, classical realist approaches are insufficient to explain the changing -and increasingly abstracted-security areas. Nation-states must develop more complex skills and prioritize cooperation to deal with these new security threats. In our study, which focuses on the national cyber security perceptions of states, the measures taken by the States in the cyber world on priority issues, such as institutional functioning and survival problems, are analyzed in terms of national security strategies. On the theme of 'new security', the main focus is the necessity for actors to develop new defense capabilities within the framework of the 'quick response and easy adaptation' in the face of increasing and diversifying cyber threats. In our study, which deals with the efforts of digital Nation-states to increase their effectiveness in cyberspace in determining the 21st-century sovereignty areas and the increase in their tendency towards cyber warfare tools, the active/passive defense methods followed by the States in the perspective of cyber security strategies have been evaluated. In this context, it is the most acceptable method for the States to prefer the 'active defense model' in order to avoid cyber attacks against vital institutions such as education and health.

An active deception defense model based on address mutation and fingerprint camouflage

An active deception defense model based on address mutation and fingerprint camouflage

Arranging the Defence of the Cyber Environment as a Part of Military Affairs: Tactical, Operational and Strategic approach in retrospect of The Russian - Ukrainian War 2022

The artificial cyber environment has reached national security interest and emerged as the fourth domain of battle in the military concept of all-domain operations in most Western armed forces and coalitions in the past 30 years. Currently, militaries are struggling to keep up with cybercriminals and advanced persistent actors while trying to gain an advantage of their data and digital infrastructure. The paper focuses on military affairs' ways and means to address the need for cyber warriors operating in friendly, neutral, and hostile cyber environments integrated under Multi-Domain Operations. The paper uses design research methodology to create and test a model for cyber defence capabilities generation and utilisation. The theoretical reference to military affairs is based on Beer's Viable System Model and previous studies of military organisations' evolution as capability generators. The military and societal cyber environment evolution model is based on industrial revolutions and current tendencies. These approaches define a hypothetical model for two main functions of military affairs (force generation and utilisation) concerning cyber defence capabilities. The fast evolution of cyber threats sets unique requirements for cyber force utilisation and generation structures. This difference has culminated in a recent war between Russia and Ukraine, and data from that conflict is used to test the hypothetical model. The rapid evolution of the cyber environment and its weaponisation establish different requirements for military cyber capabilities compared to any other operational dimension or capability (space, air, land, or maritime). The difference is evident in sourcing resources, generating capabilities, and using them in Multi-Domain Operations. The paper provides a tested model for generating cyber defence capabilities at a strategic level and an operation model for cyber defence at a tactical and operational level. The designed model extends the technically oriented cybersecurity thinking with operational and strategic levels. Furthermore, the model introduces the value stream behind the cyber capability acquisition and supports strategic designers in national and military analysis.

Modern ransomware: Evolution, methodology, attack model, prevention and mitigation using multi‐tiered approach

AbstractRansomware is a menace to the vibrant digital ecosystem. The exponential growth in ransomware attacks, its detrimental impacts, and the ever‐changing methods adopted by threat actor groups demands a focused understanding of the evolution of ransomware. This would help the organizations devise novel defensive frameworks and security controls against the modern ransomware. In this work, the impacts and evolution of ransomware through different phases up to its current form are detailed. Further, based on the study and analysis of the most prevalent modern ransomware variants, their most used tactics, techniques and procedures (TTPs) are identified as per the MITRE ATT&CK model. This acts as a platform to propose a generic attack model for “modern ransomware.” Building on the existing MITRE mitigation, D3FEND‐based approaches and considering the resource and budget constraints of organizations, a simplified three‐tier defensive model that is cost‐effective and implementable is put forward. Thus, this work aims to open avenues for understanding the TTPs, and attack methodology of “modern ransomware,” thereby developing feasible and implementable defensive security controls.

Research on defense strategies for power system frequency stability under false data injection attacks

The rapid development of cyber-physical power system has highlighted the pressing issue of cyber attacks (CAs) faced by large-scale renewable energy power systems (LREPS). The injection of false data attacks on critical electrical equipment by attackers can lead to abnormal system frequencies, triggering cascade failures and causing widespread blackouts. This paper introduces a tri-level defense model specifically crafted for LREPS, comprising an upper, middle, and lower level (defender-attacker-operator). By promoting collaborative decision-making among these agents, the model aims to mitigate the rate of change of frequency (RoCoF) and effectively withstand CAs. The upper-level employs an economic-inertia optimization strategy, scheduling generator units to maximize system inertia while minimizing generation costs. By leveraging the strong duality theorem, the tri-level model is decomposed into a bi-level model consisting of a master problem and a subproblem, solved using the Benders decomposition method. Furthermore, this paper also considers augmenting the virtual inertia of renewable energy units when defense resources are insufficient, ensuring RoCoF remains within limits after CAs. The effectiveness and superiority of the proposed model and strategy are validated through case studies based on the improved IEEE 118-bus system. The results demonstrate that the model can reduce load shedding caused by the worst-case CAs, enhance system inertia, lower RoCoF, and improve the stability of LREPS.

Breaking Down Barriers: Adapting Three Lines of Defense for Ever Changing Banking Regulations

The changing regulatory landscape presents a significant challenge to banks. This paper discusses the challenges associated with adapting the Three Lines of Defense (3LOD) model in order to comply with regulatory changes within the banking industry. The paper focuses on the challenges of adapting to the rapidly evolving banking regulatory environment. The three lines of defense (LoDs) model is a framework that enables banks to manage risk and compliance with regulatory requirements with precision and accountability. The 3LODs are based on a three-line strategy. The LLOD model enables a bank to identify and manage risks accurately and effectively. The use of technology and analytics can enable banks to automate control checks to improve the speed and precision of their control checks. To conclude, banks need to refine the three lines of defense to keep up with an ever-evolving regulatory environment to ensure compliance with rules, identifying risks accurately, and enhancing their risk management capabilities.

Interactive attack-defense for generalized person re-identification

Generalized Person Re-Identification (GReID) aims to develop a model capable of robust generalization across unseen target domains, even with training on a limited set of observed domains. Recently, methods based on the Attack-Defense mechanism are emerging as a prevailing technology to this issue, which treats domain transformation as a type of attack and enhances the model’s generalization performance on the target domain by equipping it with a defense module. However, a significant limitation of most existing approaches is their inability to effectively model complex domain transformations, largely due to the separation of attack and defense components. To overcome this limitation, we introduce an innovative Interactive Attack-Defense (IAD) mechanism for GReID. The core of IAD is the interactive learning of two models: an attack model and a defense model. The attack model dynamically generates directional attack information responsive to the current state of the defense model, while the defense model is designed to derive generalizable representations by utilizing a variety of attack samples. The training approach involves a dual process: for the attack model, the aim is to increase the challenge for the defense model in countering the attack; conversely, for the defense model, the focus is on minimizing the effects instigated by the attack model. This interactive framework allows for mutual learning between attack and defense, creating a synergistic learning environment. Our diverse experiments across datasets confirm IAD’s effectiveness, consistently surpassing current state-of-the-art methods, and using MSMT17 as the target domain in different protocols resulted in a notable 13.4% improvement in GReID task average Rank-1 accuracy. Code is available at: https://github.com/lhf12278/IAD.

Deriving optimal defensive resource deployment of complex system-of-systems for risk response considering cascading effect

As disruptions happen with increasing frequency, over-deployment or insufficient deployment can lead to inefficiency or defensive failure for security risk management. Therefore, it is necessary to deploy appropriate defensive resources so as to defend a complex system-of-systems (SoS). Component heterogeneity and cascade effect must be considered in the high-risk environment. To that end, a novel security defensive deployment decision model is proposed in this research considering cooperation and non-cooperation. We formulate a maximum utility model to figure out the optimal defensive deployment of component in the case of independent defence. In addition, the complex SoS’s optimal defensive behaviour is calculated under all components work together to fulfil the demand and complete the mission. We also compare two scenarios and draw that the utility of coordinated defence is greater than that of independent defence in a SoS. The proposed model is employed in a weapon equipment SoS, elaborating how the SoS operators should operate against disruptive events.

Fungal Necrotrophic Interaction: A Case Study of Seed Immune Response to a Seed-Borne Pathogen

Seeds play a vital role in the perpetuation of plant species, both in natural environments and agriculture. However, they often face challenges from biotic stresses, such as seed-borne pathogenic fungi. The transgenerational transmission of these seed-borne fungi, along with their dissemination during seed commercialization, can contribute to the emergence of global epidemic diseases, resulting in substantial economic losses. Despite the recognized impact of seed-borne pathogens on agriculture, our understanding of seed–pathogen interactions remains limited. This review establishes parallels between the current state of knowledge regarding seed responses to pathogen interactions and well-established plant defense models, primarily derived from typical physiological conditions observed during leaf infections. Examining fragmented results from various pathosystems, this review seeks to offer a comprehensive overview of our current understanding of interactions during seed development and germination. The necrotrophic interactions in Brassicaceae are described using recent transcriptomic and genetic studies focused on the Arabidopsis/Alternaria pathosystem, which illustrates original response pathways in germinating seeds that markedly differ from the general concept of plant–pathogen interactions. The co-existence of regulatory mechanisms affecting both seed resistance and susceptibility, potentially promoting fungal colonization, is examined. The vulnerable response during germination emerges as a crucial consideration in the context of sustainable plant health management in agriculture.

LRS: Enhancing Adversarial Transferability through Lipschitz Regularized Surrogate

The transferability of adversarial examples is of central importance to transfer-based black-box adversarial attacks. Previous works for generating transferable adversarial examples focus on attacking given pretrained surrogate models while the connections between surrogate models and adversarial trasferability have been overlooked. In this paper, we propose Lipschitz Regularized Surrogate (LRS) for transfer-based black-box attacks, a novel approach that transforms surrogate models towards favorable adversarial transferability. Using such transformed surrogate models, any existing transfer-based black-box attack can run without any change, yet achieving much better performance. Specifically, we impose Lipschitz regularization on the loss landscape of surrogate models to enable a smoother and more controlled optimization process for generating more transferable adversarial examples. In addition, this paper also sheds light on the connection between the inner properties of surrogate models and adversarial transferability, where three factors are identified: smaller local Lipschitz constant, smoother loss landscape, and stronger adversarial robustness. We evaluate our proposed LRS approach by attacking state-of-the-art standard deep neural networks and defense models. The results demonstrate significant improvement on the attack success rates and transferability. Our code is available at https://github.com/TrustAIoT/LRS.

Defending against model poisoning attack in federated learning: A variance-minimization approach

The distributed nature of federated learning (FL) renders the learning process susceptible to model poisoning attacks, whereby local workers in FL report fabricated and false local training outcomes to the FL server with the intention to compromise/degrade the global model, or even derail the global learning process so that it can no longer converge. Existing defense mechanisms typically consider the falsified local updates as outliers that reside far away from the mean update, and attempt to counter against such model poisoning attacks by detecting and eliminating those outliers in the reported local training updates. These methods do not perform well when the data of different workers are non-I.I.D. and/or when there are multiple colluding attackers, under which an outlier update is not always a falsified update, and vice versa. In this paper, we propose a novel defense mechanism, MinVar, to counter the model poisoning attacks in FL from a drastically different perspective. Instead of detecting and eliminating outlier local updates from the global model aggregation, MinVar takes all local updates but assigns different weights to them in the global model aggregation. MinVar decides the optimal weights by formulating and solving an optimization problem in each iteration of the learning process, which aims to suppress the contribution of those falsified (i.e., malicious) updates while still retaining the contribution of those honest/truthful updates. Based on the sparsity observation in most deep neural networks, a data sampling technique is further proposed to reduce the computation complexity of MinVar while preserving its defense performance. Extensive experiments are conducted on both the MNIST and CIFAR-10 datasets, and the results verify the effectiveness of the proposed MinVar defense model.

RDMAA: Robust Defense Model against Adversarial Attacks in Deep Learning for Cancer Diagnosis

RDMAA: Robust Defense Model against Adversarial Attacks in Deep Learning for Cancer Diagnosis