arXiv subjects

Shouhuai Xu

Publications and source records attributed to Shouhuai Xu.

At least 37 records · Page 2Linked to original sources

arXiv2025

Game-Theoretic Cybersecurity: the Good, the Bad and the Ugly

Given the scale of consequences attributable to cyber attacks, the field of cybersecurity has long outgrown ad-hoc decision-making. A popular choice to provide disciplined decision-making in cybersecurity is Game Theory, which seeks to mathematically understand strategic interaction. In practice though, game-theoretic approaches are scarcely utilized (to our knowledge), highlighting the need to understand the deficit between the existing state-of-the-art and the needs of cybersecurity practitioners. Therefore, we develop a framework to characterize the function and assumptions of existing works as applied to cybersecurity and leverage it to characterize 80 unique technical papers. Then, we leverage this information to analyze the capabilities of the proposed models in comparison to the application-specific needs they are meant to serve, as well as the practicality of implementing the proposed solution. Our main finding is that Game Theory largely fails to incorporate notions of uncertainty critical to the application being considered. To remedy this, we provide guidance in terms of how to incorporate uncertainty in a model, what forms of uncertainty are critical to consider in each application area, and how to model the information that is available in each application area.

cs.GT↗

arXiv2024

AutoCRAT: Automatic Cumulative Reconstruction of Alert Trees

When a network is attacked, cyber defenders need to precisely identify which systems (i.e., computers or devices) were compromised and what damage may have been inflicted. This process is sometimes referred to as cyber triage and is an important part of the incident response procedure. Cyber triage is challenging because the impacts of a network breach can be far-reaching with unpredictable consequences. This highlights the importance of automating this process. In this paper we propose AutoCRAT, a system for quantifying the breadth and severity of threats posed by a network exposure, and for prioritizing cyber triage activities during incident response. Specifically, AutoCRAT automatically reconstructs what we call alert trees, which track network security events emanating from, or leading to, a particular computer on the network. We validate the usefulness of AutoCRAT using a real-world dataset. Experimental results show that our prototype system can reconstruct alert trees efficiently and can facilitate data visualization in both incident response and threat intelligence analysis.

cs.CR↗

arXiv2024

Smart Home Cyber Insurance Pricing

Our homes are increasingly employing various kinds of Internet of Things (IoT) devices, leading to the notion of smart homes. While this trend brings convenience to our daily life, it also introduces cyber risks. To mitigate such risks, the demand for smart home cyber insurance has been growing rapidly. However, there are no studies on analyzing the competency of smart home cyber insurance policies offered by cyber insurance vendors (i.e., insurers), where `competency' means the insurer is profitable and smart home owners are not overly charged with premiums and/or deductibles. In this paper, we propose a novel framework for pricing smart home cyber insurance, which can be adopted by insurers in practice. Our case studies show, among other things, that insurers are over charging smart home owners in terms of premiums and deductibles.

cs.CE↗

arXiv2024

Quantifying Psychological Sophistication of Malicious Emails

Malicious emails including Phishing, Spam, and Scam are one significant class of cyber social engineering attacks. Despite numerous defenses to counter them, the problem remains largely open. The ineffectiveness of current defenses can be attributed to our superficial understanding of the psychological properties that make these attacks successful. This problem motivates us to investigate the psychological sophistication, or sophistication for short, of malicious emails. We propose an innovative framework that accommodates two important and complementary aspects of sophistication, dubbed Psychological Techniques, PTechs, and Psychological Tactics, PTacs. We propose metrics and grading rules for human experts to assess the sophistication of malicious emails via the lens of these PTechs and PTacs. To demonstrate the usefulness of the framework, we conduct a case study based on 1,036 malicious emails assessed by four independent graders. Our results show that malicious emails are psychologically sophisticated, while exhibiting both commonalities and different patterns in terms of their PTechs and PTacs. Results also show that previous studies might have focused on dealing with the less proliferated PTechs such as Persuasion and PTacs such as Reward, rather than the most proliferated PTechs such as Attention Grabbing and Impersonation, and PTacs such as Fit and Form and Familiarity that are identified in this study. We also found among others that social events are widely exploited by attackers in contextualizing their malicious emails. These findings could be leveraged to guide the design of effective defenses against malicious emails.

cs.CR↗

arXiv2024

Characterizing the Evolution of Psychological Factors Exploited by Malicious Emails

Cyber attacks, including cyber social engineering attacks, such as malicious emails, are always evolving with time. Thus, it is important to understand their evolution. In this paper we characterize the evolution of malicious emails through the lens of Psychological Factors, PFs, which are humans psychological attributes that can be exploited by malicious emails. That is, attackers who send them. For this purpose, we propose a methodology and apply it to conduct a case study on 1,260 malicious emails over a span of 21 years, 2004 to 2024. Our findings include attackers have been constantly seeking to exploit many PFs, especially the ones that reflect human traits. Attackers have been increasingly exploiting 9 PFs and mostly in an implicit or stealthy fashion. Some PFs are often exploited together. These insights shed light on how to design future defenses against malicious emails.

cs.CR↗

arXiv2024

Characterizing the Evolution of Psychological Tactics and Techniques Exploited by Malicious Emails

The landscape of malicious emails and cyber social engineering attacks in general are constantly evolving. In order to design effective defenses against these attacks, we must deeply understand the Psychological Tactics, PTacs, and Psychological Techniques, PTechs, that are exploited by these attacks. In this paper we present a methodology for characterizing the evolution of PTacs and PTechs exploited by malicious emails. As a case study, we apply the methodology to a real-world dataset. This leads to a number insights, such as which PTacs or PTechs are more often exploited than others. These insights shed light on directions for future research towards designing psychologically-principled solutions to effectively counter malicious emails.

cs.CR↗

arXiv2024

Towards Principled Risk Scores for Space Cyber Risk Management

Space is an emerging domain critical to humankind. Correspondingly, space cybersecurity is an emerging field with much research to be done. To help space cybersecurity practitioners better manage cyber risks, The Aerospace Corporation proposed Notional Risk Scores (NRS) within their Space Attack Research and Tactic Analysis (SPARTA) framework, which can be applied to quantify the cyber risks associated with space infrastructures and systems. While intended for adoption by practitioners, NRS has not been analyzed with real-world scenarios, putting its effectiveness into question. In this paper we analyze NRS via a real-world cyber attack scenario against a satellite, and characterize the strengths, weaknesses, and applicability of NRS. The characterization prompts us to propose a set of desired properties to guide the design of future NRS. As a first step along this direction, we further propose a formalism to serve as a baseline for designing future NRS with those desired properties.

cs.CR↗

arXiv2024

On the Effectiveness of Function-Level Vulnerability Detectors for Inter-Procedural Vulnerabilities

Software vulnerabilities are a major cyber threat and it is important to detect them. One important approach to detecting vulnerabilities is to use deep learning while treating a program function as a whole, known as function-level vulnerability detectors. However, the limitation of this approach is not understood. In this paper, we investigate its limitation in detecting one class of vulnerabilities known as inter-procedural vulnerabilities, where the to-be-patched statements and the vulnerability-triggering statements belong to different functions. For this purpose, we create the first Inter-Procedural Vulnerability Dataset (InterPVD) based on C/C++ open-source software, and we propose a tool dubbed VulTrigger for identifying vulnerability-triggering statements across functions. Experimental results show that VulTrigger can effectively identify vulnerability-triggering statements and inter-procedural vulnerabilities. Our findings include: (i) inter-procedural vulnerabilities are prevalent with an average of 2.8 inter-procedural layers; and (ii) function-level vulnerability detectors are much less effective in detecting to-be-patched functions of inter-procedural vulnerabilities than detecting their counterparts of intra-procedural vulnerabilities.

cs.CR↗

arXiv2024

Optimally Blending Honeypots into Production Networks: Hardness and Algorithms

Honeypot is an important cyber defense technique that can expose attackers new attacks. However, the effectiveness of honeypots has not been systematically investigated, beyond the rule of thumb that their effectiveness depends on how they are deployed. In this paper, we initiate a systematic study on characterizing the cybersecurity effectiveness of a new paradigm of deploying honeypots: blending honeypot computers (or IP addresses) into production computers. This leads to the following Honeypot Deployment (HD) problem, How should the defender blend honeypot computers into production computers to maximize the utility in forcing attackers to expose their new attacks while minimizing the loss to the defender in terms of the digital assets stored in the compromised production computers? We formalize HD as a combinatorial optimization problem, prove its NP hardness, provide a near optimal algorithm (i.e., polynomial time approximation scheme). We also conduct simulations to show the impact of attacker capabilities.

cs.CR↗

arXiv2023

Robin: A Novel Method to Produce Robust Interpreters for Deep Learning-Based Code Classifiers

Deep learning has been widely used in source code classification tasks, such as code classification according to their functionalities, code authorship attribution, and vulnerability detection. Unfortunately, the black-box nature of deep learning makes it hard to interpret and understand why a classifier (i.e., classification model) makes a particular prediction on a given example. This lack of interpretability (or explainability) might have hindered their adoption by practitioners because it is not clear when they should or should not trust a classifier's prediction. The lack of interpretability has motivated a number of studies in recent years. However, existing methods are neither robust nor able to cope with out-of-distribution examples. In this paper, we propose a novel method to produce \underline{Rob}ust \underline{in}terpreters for a given deep learning-based code classifier; the method is dubbed Robin. The key idea behind Robin is a novel hybrid structure combining an interpreter and two approximators, while leveraging the ideas of adversarial training and data augmentation. Experimental results show that on average the interpreter produced by Robin achieves a 6.11\% higher fidelity (evaluated on the classifier), 67.22\% higher fidelity (evaluated on the approximator), and 15.87x higher robustness than that of the three existing interpreters we evaluated. Moreover, the interpreter is 47.31\% less affected by out-of-distribution examples than that of LEMNA.

cs.SE↗

arXiv2023

Characterizing Cyber Attacks against Space Systems with Missing Data: Framework and Case Study

Cybersecurity of space systems is an emerging topic, but there is no single dataset that documents cyber attacks against space systems that have occurred in the past. These incidents are often scattered in media reports while missing many details, which we dub the missing-data problem. Nevertheless, even "low-quality" datasets containing such reports would be extremely valuable because of the dearth of space cybersecurity data and the sensitivity of space systems which are often restricted from disclosure by governments. This prompts a research question: How can we characterize real-world cyber attacks against space systems? In this paper, we address the problem by proposing a framework, including metrics, while also addressing the missing-data problem, by "extrapolating" the missing data in a principled fashion. To show the usefulness of the framework, we extract data for 72 cyber attacks against space systems and show how to extrapolate this "low-quality" dataset to derive 4,076 attack technique kill chains. Our findings include: cyber attacks against space systems are getting increasingly sophisticated; and, successful protection against on-path and social engineering attacks could have prevented 80% of the attacks.

cs.CR↗

arXiv2023

Towards Automated Cyber Range Design: Characterizing and Matching Demands to Supplies

Cyber ranges mimic real-world cyber environments and are in high demand. Before building their own cyber ranges, organizations need to deeply understand what construction supplies are available to them. A fundamental supply is the cyber range architecture, which prompts an important research question: Which cyber range architecture is most appropriate for an organization's requirements? To answer this question, we propose an innovative framework to specify cyber range requirements, characterize cyber range architectures (based on our analysis of 45 cyber range architectures), and match cyber range architectures to cyber range requirements.

cs.CR↗

arXiv2023

Analysis of Contagion Dynamics with Active Cyber Defenders

In this paper, we analyze the infection spreading dynamics of malware in a population of cyber nodes (i.e., computers or devices). Unlike most prior studies where nodes are reactive to infections, in our setting some nodes are active defenders meaning that they are able to clean up malware infections of their neighboring nodes, much like how spreading malware exploits the network connectivity properties in order to propagate. We formulate these dynamics as an Active Susceptible-Infected-Susceptible (A-SIS) compartmental model of contagion. We completely characterize the system's asymptotic behavior by establishing conditions for the global asymptotic stability of the infection-free equilibrium and for an endemic equilibrium state. We show that the presence of active defenders counter-acts infectious spreading, effectively increasing the epidemic threshold on parameters for which an endemic state prevails. Leveraging this characterization, we investigate a general class of problems for finding optimal investments in active cyber defense capabilities given limited resources. We show that this class of problems has unique solutions under mild assumptions. We then analyze an Active Susceptible-Infected-Recovered (A-SIR) compartmental model, where the peak infection level of any trajectory is explicitly derived.

eess.SY↗

arXiv2023

PAD: Towards Principled Adversarial Malware Detection Against Evasion Attacks

Machine Learning (ML) techniques can facilitate the automation of malicious software (malware for short) detection, but suffer from evasion attacks. Many studies counter such attacks in heuristic manners, lacking theoretical guarantees and defense effectiveness. In this paper, we propose a new adversarial training framework, termed Principled Adversarial Malware Detection (PAD), which offers convergence guarantees for robust optimization methods. PAD lays on a learnable convex measurement that quantifies distribution-wise discrete perturbations to protect malware detectors from adversaries, whereby for smooth detectors, adversarial training can be performed with theoretical treatments. To promote defense effectiveness, we propose a new mixture of attacks to instantiate PAD to enhance deep neural network-based measurements and malware detectors. Experimental results on two Android malware datasets demonstrate: (i) the proposed method significantly outperforms the state-of-the-art defenses; (ii) it can harden ML-based malware detection against 27 evasion attacks with detection accuracies greater than 83.45%, at the price of suffering an accuracy decrease smaller than 2.16% in the absence of attacks; (iii) it matches or outperforms many anti-malware scanners in VirusTotal against realistic adversarial malware.

cs.CR↗

arXiv2022

Internet-based Social Engineering Attacks, Defenses and Psychology: A Survey

Social engineering attacks are a major cyber threat because they often serve as a first step for an attacker to break into an otherwise well-defended network, steal victims' credentials, and cause financial losses. The problem has received due amount of attention with many publications proposing defenses against them. Despite this, the situation has not improved. In this paper, we aim to understand and explain this phenomenon by looking into the root cause of the problem. To this end, we examine the literature on attacks and defenses through a unique lens we propose -- {\em psychological factors (PFs) and techniques (PTs)}. We find that there is a big discrepancy between attacks and defenses: Attacks have deliberately exploited PFs by leveraging PTs, but defenses rarely take either of these into consideration, preferring technical solutions. This explains why existing defenses have achieved limited success. This prompts us to propose a roadmap for a more systematic approach towards designing effective defenses against social engineering attacks.

cs.CR↗

arXiv2022

A Coupling Approach to Analyzing Games with Dynamic Environments

The theory of learning in games has extensively studied situations where agents respond dynamically to each other by optimizing a fixed utility function. However, in real situations, the strategic environment varies as a result of past agent choices. Unfortunately, the analysis techniques that enabled a rich characterization of the emergent behavior in static environment games fail to cope with dynamic environment games. To address this, we develop a general framework using probabilistic couplings to extend the analysis of static environment games to dynamic ones. Using this approach, we obtain sufficient conditions under which traditional characterizations of Nash equilibria with best response dynamics and stochastic stability with log-linear learning can be extended to dynamic environment games. As a case study, we pose a model of cyber threat intelligence sharing between firms and a simple dynamic game-theoretic model of social precautions in an epidemic, both of which feature dynamic environments. For both examples, we obtain conditions under which the emergent behavior is characterized in the dynamic game by performing the traditional analysis on a reference static environment game.

cs.GT↗

arXiv2022

RoPGen: Towards Robust Code Authorship Attribution via Automatic Coding Style Transformation

Source code authorship attribution is an important problem often encountered in applications such as software forensics, bug fixing, and software quality analysis. Recent studies show that current source code authorship attribution methods can be compromised by attackers exploiting adversarial examples and coding style manipulation. This calls for robust solutions to the problem of code authorship attribution. In this paper, we initiate the study on making Deep Learning (DL)-based code authorship attribution robust. We propose an innovative framework called Robust coding style Patterns Generation (RoPGen), which essentially learns authors' unique coding style patterns that are hard for attackers to manipulate or imitate. The key idea is to combine data augmentation and gradient augmentation at the adversarial training phase. This effectively increases the diversity of training examples, generates meaningful perturbations to gradients of deep neural networks, and learns diversified representations of coding styles. We evaluate the effectiveness of RoPGen using four datasets of programs written in C, C++, and Java. Experimental results show that RoPGen can significantly improve the robustness of DL-based code authorship attribution, by respectively reducing 22.8% and 41.0% of the success rate of targeted and untargeted attacks on average.

cs.CR↗

arXiv2021

Quantifying Cybersecurity Effectiveness of Dynamic Network Diversity

The deployment of monoculture software stacks can have devastating consequences because a single attack can compromise all of the vulnerable computers in cyberspace. This one-vulnerability-affects-all phenomenon will continue until after software stacks are diversified, which is well recognized by the research community. However, existing studies mainly focused on investigating the effectiveness of software diversity at the building-block level (e.g., whether two independent implementations indeed exhibit independent vulnerabilities); the effectiveness of enforcing network-wide software diversity is little understood, despite its importance in possibly helping justify investment in software diversification. As a first step towards ultimately tackling this problem, we propose a systematic framework for modeling and quantifying the cybersecurity effectiveness of network diversity, including a suite of cybersecurity metrics. We also present an agent-based simulation to empirically demonstrate the usefulness of the framework. We draw a number of insights, including the surprising result that proactive diversity is effective under very special circumstances, but reactive-adaptive diversity is much more effective in most cases.

cs.CR↗