Google Safe Browsing (GSB) and DNS resolution operate concurrently during browser navigation, yet their packet-level synchronization remains understudied. This work characterizes the timing gap (Δ_time) between GSB-related query close events and parallel DNS resolution responses, identifying a consistent temporal offset with potential security relevance. Using packet-capture analysis across general and CNAME-domain datasets, we observe positive timing gaps in approximately 79% of measurements. In these instances, DNS responses lag behind GSB-related query closures with median delays of 67-79 ms and maximum delays surpassing 2,400 ms. These empirical results highlight a measurable window within the browsing workflow. We suggest that such temporal inconsistencies, particularly in complex CNAME-domain resolutions, may create a security-relevant timing precondition under DNS-manipulation threat models. These results provide a foundation for further research into timing-based risks and mitigations in browser safety mechanisms.

In the PKI system, revoking problematic certificates in a timely manner is essential for secure communications on the Internet. Regrettably, the primary revocation schemes have shown limitations in various aspects such as privacy and scalability. Although many state-of-the-art revocation techniques have been proposed to overcome such limitations, the implications of their deployment in practice in each aspect have not been deeply studied. This article aims to systematically classify and comparatively analyze the existing schemes in terms of security, efficiency, and practicality. Based on our evaluation, we also provide insight into securing the revocation systems, and finally discuss their future directions.

State-of-the-art reference-based phishing detectors using Siamese networks are vulnerable to adversarial attacks. Existing methods, such as LogoMorph, however, suffer from a trade-off between evasion effectiveness and visual fidelity, often producing conspicuous artifacts due to the stochastic nature of diffusion models. We present LogoFlow, an attention-guided adversarial generator built upon a deterministic Rectified Flow backbone. By leveraging GradCAM to identify critical regions, LogoFlow adaptively suppresses perturbations in high-attention areas to preserve identity while amplifying them in low-attention zones to disrupt similarity. Against PhishIntention, LogoFlow improves logo-level evasion by up to 83 percentage points over Logo-Morph and achieves an order-of-magnitude improvement in webpage-level evasion (e.g., 92.2% vs. 9.2% on Google). A user study confirms that participants failed to identify 80% of LogoFlow -generated phishing pages, exposing critical robustness gaps in Siamese-based detectors.

While characterizing network connections in their early stage is vital for providing timely responses against network threats, existing methods become less attractive due to the requirement of complete connection information (thus unable to make timely identification) or packet payload inspection (therefore limited to unencrypted packets under no privacy regulation). To this end, this paper takes an approach of packet stream analysis referencing statistical information of packet sequences, requiring neither packet inspection nor complete connection information. To enable practical packet stream analysis, there exist several challenges, such as out-of-order packet sequences introduced by network dynamics and class imbalance with a tiny fraction of attack connections. To overcome these challenges, we design two deep sequence models: 1) a bidirectional recurrent structure designed for greater resilience to out-of-order packet streams; and 2) a pre-training-enabled sequence-to-sequence structure designed for creating consistent representations from unbalanced class distributions using self-supervised learning. We evaluate the presented deep sequence models using real and synthetic network data collections for extensive experimentation. The experimental results support the feasibility of the proposed models outperforming baseline deep learning models, yielding up to 94.8% (F1 score) only with the first five packets ( k =5) from the Internet traffic collection containing a substantial fraction of network flows experiencing out-of-order delivery.

Large language model (LLM)-powered content moderation systems have become a critical defense against harmful online content. However, these systems primarily operate on tokenized text and largely ignore the visual cues that humans naturally rely on when interpreting content. We show that this discrepancy creates a fundamental perceptual mismatch: content that is readily recognized as harmful by humans can become effectively invisible to automated moderation systems. To study this vulnerability, we introduce a class of Human-Perceptible Adversarial Attacks (HPAA), in which harmful expressions are embedded into otherwise benign text through visually salient typographic manipulations. Our key insight is that typographic features, including spacing, visual emphasis, and spatial arrangement, can be strategically combined to preserve human recognition of harmful content while substantially reducing machine detectability. Operating in black-box settings with only a small query budget, our attack automatically generates evasive content without requiring model access or gradient information. We evaluate the attack across multiple datasets and ten deployed moderation systems, including commercial APIs and state-of-the-art open-source guardrails. Results reveal a striking gap between human and machine perception: with only three detector queries, generated attacks achieve over 86% human recognition while maintaining detection rates below 1% across the evaluated systems. We further conduct ablation studies to identify the typographic factors driving successful evasion, analyze why current moderation architectures fail to capture these signals, and discuss practical defenses. Our findings expose a fundamental blind spot in today's LLM-based moderation ecosystem and highlight need for moderation systems that reason about content in a manner more consistent with human perceptual understanding.

Code-generating Large Language Models (LLMs) significantly accelerate software development. However, their frequent generation of insecure code presents serious risks. We present a comprehensive evaluation of seven parameter-efficient fine-tuning (PEFT) techniques, demonstrating substantial gains in secure code generation without compromising functionality. Our research identifies prompt-tuning as the most effective PEFT method, achieving an 80.86

Despite extensive research on text passwords, the security and memorability of verbal passwords-spoken rather than typed-remain underexplored. Verbal passwords hold significant potential for scenarios where keyboard input is impractical (e.g., smart speakers, wearables, vehicles) or users have motor impairments that make typing difficult. Through two large-scale user studies, we assessed the viability of verbal passwords. In our first study (N = 2,085), freely chosen verbal passwords were found to have a limited guessing space, with 39.76% cracked within 109 guesses. However, in our second study (n = 600), applying word count and blocklist policies for verbal password creation significantly enhanced verbal password performance, achieving better memorability and security than traditional text passwords. Specifically, 65.6% of verbal password users (under the password creation policy using minimum word counts and a blocklist) successfully recalled their passwords in long-term tests, compared to 54.11% for text passwords. Additionally, verbal passwords with enforced policies exhibited a lower crack rate (6.5%) than text passwords (10.3%). These findings highlight verbal passwords as a practical and secure alternative for contexts where text passwords are infeasible, offering strong memorability with robust resistance to guessing attacks.

Phishing attacks pose a significant threat to Internet users. Understanding the security posture of phishing infrastructure is crucial for developing effective defense strategies, as it helps identify potential weaknesses that attackers might exploit. Despite extensive research, there may still be a gap in fully understanding these security weaknesses. To address this important issue, this paper presents a longitudinal study of security configurations and vulnerabilities in phishing websites and associated kits. We focus on two main areas: (1) analyzing the security configurations of phishing websites and servers, particularly HTTP headers and application-level security, and (2) examining the prevalence and types of vulnerabilities in phishing kits. We analyze data from 906,731 distinct phishing websites collected over 2.5 years, covering HTML headers, client-side resources, and phishing kits. Our findings suggest that phishing websites often employ weak security configurations, with 88.8% of the 13,344 collected phishing kits containing at least one potential vulnerability, and 12.5% containing backdoor vulnerabilities. These vulnerabilities present an opportunity for defenders to shift from passive defense to active disruption of phishing operations. Our research proposes a new approach to leverage weaknesses in phishing infrastructure, allowing defenders to take proactive actions to disable phishing sites earlier and reduce their effectiveness.

Malware often employs encryption to obfuscate its network communications, posing challenges for network-based anomaly detection techniques. Distinguishing encrypted packets from one another becomes particularly difficult, especially when operating under zero knowledge setting, such as detecting new malware. Existing approaches rely on unique features extracted from network connections to train deep learning algorithms. However, these methods fall short when dealing with new malware due to limited information. To address this challenge, we propose Anti-EMP, an algorithm designed to filter encrypted malware packets. Specifically, Anti-EMP identifies and sifts out encrypted packets originating from the same malware across multiple clients. Our approach is grounded in two practical assumptions: (a) the packets were encrypted using an unknown, identical stream cipher and encryption key and (b) a suspicious packet related to malware can be captured. We also propose a novel method for generating Anti-EMP, significantly enhancing the capability to detect encrypted malware packets without prior knowledge, i.e., zero knowledge settings. Our experiments show that Anti-EMP can be generated in approximately one second, facilitating easy iteration and easy selection of another suspicious packet if it turns out to be ineffective. Notably, our proposed scheme demonstrates high effectiveness, achieving a True-Positive Rate (TPR) of 0.998 and a False-Positive Rate (FPR) of 0.001.

Large language models (LLMs) have emerged as a promising phishing detection mechanism, addressing the limitations of traditional deep learning-based detectors, including poor generalization to previously unseen websites and a lack of interpretability. However, LLMs' effectiveness for phishing detection remains unexplored. This study investigates how to effectively leverage LLMs for phishing detection (including target brand identification) by examining the impact of input modalities (screenshots, logos, HTML, and URLs), temperature settings, and prompt engineering strategies. Using a dataset of 19,131 real-world phishing websites and 243 benign sites, we evaluate seven LLMs -- two commercial models (GPT 4.1 and Gemini 2.0 flash) and five open-source models (Qwen, Llama, Janus, DeepSeek-VL2, and R1) -- alongside two deep learning (DL)-based baselines (PhishIntention and Phishpedia). Our findings reveal that commercial LLMs generally outperform open-source models in phishing detection, while DL models demonstrate better performance on benign samples. For brand identification, screenshot inputs achieve optimal results, with commercial LLMs reaching 93-95% accuracy and open-source models, particularly Qwen, achieving up to 92%. However, incorporating multiple input modalities simultaneously or applying one-shot prompts does not consistently enhance performance and may degrade results. Furthermore, higher temperature values reduce performance. Based on these results, we recommend using screenshot inputs with zero temperature to maximize accuracy for LLM-based detectors with HTML serving as auxiliary context when screenshot information is insufficient.

Phishing attacks pose a significant threat to Internet users, with cybercriminals elaborately replicating the visual appearance of legitimate websites to deceive victims. Visual similarity-based detection systems have emerged as an effective countermeasure, but their effectiveness and robustness in real-world scenarios have been underexplored. In this paper, we comprehensively scrutinize and evaluate the effectiveness and robustness of popular visual similarity-based anti-phishing models using a large-scale dataset of 451k real-world phishing websites. Our analyses of the effectiveness reveal that while certain visual similarity-based models achieve high accuracy on curated datasets in the experimental settings, they exhibit notably low performance on real-world datasets, highlighting the importance of real-world evaluation. Furthermore, we find that the attackers evade the detectors mainly in three ways: (1) directly attacking the model pipelines, (2) mimicking benign logos, and (3) employing relatively simple strategies such as eliminating logos from screenshots. To statistically assess the resilience and robustness of existing models against adversarial attacks, we categorize the strategies attackers employ into visible and perturbation-based manipulations and apply them to website logos. We then evaluate the models' robustness using these adversarial samples. Our findings reveal potential vulnerabilities in several models, emphasizing the need for more robust visual similarity techniques capable of withstanding sophisticated evasion attempts. We provide actionable insights for enhancing the security of phishing defense systems, encouraging proactive actions.

An advanced persistent threat (APT) refers to a covert, long-term cyberattack, typically conducted by state-sponsored actors, targeting critical sectors and often remaining undetected for long periods. In response, collective intelligence from around the globe collaborates to identify and trace surreptitious activities, generating substantial documentation on APT campaigns publicly available on the web. While prior works predominantly focus on specific aspects of APT cases, such as detection, evaluation, cyber threat intelligence, and dataset creation, limited attention has been devoted to revisiting and investigating these scattered dossiers in a longitudinal manner. The objective of our study is to fill the gap by offering a macro perspective, connecting key insights and global trends in past APT attacks. We systematically analyze six reliable sources-three focused on technical reports and another three on threat actors-examining 1,509 APT dossiers (24,215 pages) spanning 2014-2023, and identifying 603 unique APT groups worldwide. To efficiently unearth relevant information, we employ a hybrid methodology that combines rule-based information retrieval with large-language-model-based search techniques. Our longitudinal analysis reveals shifts in threat actor activities, global attack vectors, changes in targeted sectors, and relationships between cyberattacks and significant events such as elections or wars, which provide insights into historical patterns in APT evolution. Over the past decade, 154 countries have been affected, primarily using malicious documents and spear phishing as dominant initial infiltration vectors, with a noticeable decline in zero-day exploitation since 2016. Furthermore, we present our findings through interactive visualization tools, such as an APT map or flow diagram, to facilitate intuitive understanding of global patterns and trends in APT activities.

The advent of Artificial Intelligence (AI), particularly large language models (LLMs), has revolutionized software development by enabling developers to specify tasks in natural language and receive corresponding code, boosting productivity. However, this shift also introduces security risks, as LLMs may generate insecure code that can be exploited by adversaries. Conventional educational approaches emphasize efficiency while overlooking these risks, leaving students unprepared to identify and mitigate security issues in AI-assisted workflows. To surface this gap, we present , a classroom measurement and feedback framework that pairs an adversarially configured code-generation model with a VS Code extension and automated vulnerability reports that instructors can use to guide follow-up discussions. Through classroom deployments with undergraduate students (n=61), we observe that students frequently accepted insecure LLM-generated code despite prior security coursework and stated skepticism. A post-feedback survey (n=21) provides preliminary evidence that students' stated trust shifted toward greater skepticism after receiving Bifröst feedback, and that some students articulated more security-specific concerns about AI-generated code.

As datasets grow, users increasingly rely on cloud services for data storage and processing. Consequently, concerns regarding data protection and the practical use of encrypted data have emerged as significant challenges. One promising solution is order-revealing encryption (ORE), which enables efficient operations on encrypted numerical data. To support distributed environments with different users, delegatable ORE (DORE) extends this functionality to multi-client settings, enabling order comparisons between ciphertexts encrypted under different secret keys. However, Hahn et al. proposed a token forgery attack against DORE with a threat model and introduced the secure DORE (SEDORE) scheme as a countermeasure. Despite this enhancement, we claim that SEDORE remains vulnerable under the same threat model. In this paper, we present a novel Universal Token Reusability Attack, which exposes a critical vulnerability in SEDORE with the identical threat model. To mitigate this, we introduce the concept of verifiable delegatable order-revealing encryption (VDORE), along with a formal definition of token unforgeability. Building on this, we design a new scheme, Token Unforgeable DORE ( ), which ensures token unforgeability. Moreover, achieves 1.5 × faster token generation than SEDORE with enhanced security.

Phishing continues to pose a significant cybersecurity threat. While blocklists currently serve as a primary defense, due to their reactive, passive nature, these delayed responses leave phishing websites operational long enough to harm potential victims. It is essential to address this fundamental challenge at the root, particularly in phishing domains. Domain registration presents a crucial intervention point, as domains serve as the primary gateway between users and websites. We conduct a comprehensive longitudinal analysis of 690,502 unique phishing domains, spanning a 39 month period, to examine their characteristics and behavioral patterns throughout their lifecycle-from initial registration to detection and eventual deregistration. We find that 66.1 the domains in our dataset are maliciously registered, leveraging cost-effective TLDs and targeting brands by mimicking their domain names under alternative TLDs (e.g., .top and .tk) instead of the TLDs under which the brand domains are registered (e.g., .com and .ru). We also observe minimal improvements in detection speed for maliciously registered domains compared to compromised domains. Detection times vary widely across blocklists, and phishing domains remain accessible for an average of 11.5 days after detection, prolonging their potential impact. Our systematic investigation uncovers key patterns from registration through detection to deregistration, which could be leveraged to enhance anti-phishing active defenses at the DNS level.

Phishing attacks continue to be a major threat to internet users, causing data breaches, financial losses, and identity theft. This study provides an in-depth analysis of the lifespan and evolution of phishing websites, focusing on their survival strategies and evasion techniques. We analyze 286,237 unique phishing URLs over five months using a custom web crawler based on Puppeteer and Chromium. Our crawler runs on a 30-minute cycle, systematically checking the operational status of phishing websites by collecting their HTTP status codes, screenshots, HTML, and HTTP data. Temporal and survival analyses, along with statistical tests, are used to examine phishing website lifecycles, evolution, and evasion tactics. Our findings show that the average lifespan of phishing websites is 54 hours (2.25 days) with a median of 5.46 hours, indicating rapid takedown of many sites while a subset remains active longer. Interestingly, logistic-themed phishing websites (e.g., USPS) operate within a compressed timeframe (1.76 hours) compared to other brands (e.g., Facebook). We further analyze detection effectiveness using Google Safe Browsing (GSB). We find that GSB detects only 18.4% of phishing websites, taking an average of 4.5 days. Notably, 83.93% of phishing sites are already taken down before GSB detection, meaning GSB requires more prompt detection. Moreover, 16.07% of phishing sites persist beyond this point, surviving for an additional 7.2 days on average, resulting in an average total lifespan of approximately 12 days. We reveal that DNS resolution error is the main cause (67%) of phishing website takedowns. Finally, we uncover that phishing sites with extensive visual changes (more than 100 times) exhibit a median lifespan of 17 days, compared to 1.93 hours for those with minimal modifications. These results highlight the dynamic nature of phishing attacks, the challenges in detection and prevention, and the need for more rapid and comprehensive countermeasures against evolving phishing tactics.

Large Language Models (LLMs) have transformed code completion tasks, providing context-based suggestions to boost developer productivity in software engineering. As users often fine-tune these models for specific applications, poisoning and backdoor attacks can covertly alter the model outputs. To address this critical security challenge, we introduce CodeBreaker, a pioneering LLM-assisted backdoor attack framework on code completion models. Unlike recent attacks that embed malicious payloads in detectable or irrelevant sections of the code (e.g., comments), CodeBreaker leverages LLMs (e.g., GPT-4) for sophisticated payload transformation (without affecting functionalities), ensuring that both the poisoned data for fine-tuning and generated code can evade strong vulnerability detection. CodeBreaker stands out with its comprehensive coverage of vulnerabilities, making it the first to provide such an extensive set for evaluation. Our extensive experimental evaluations and user studies underline the strong attack performance of CodeBreaker across various settings, validating its superiority over existing approaches. By integrating malicious payloads directly into the source code with minimal transformation, CodeBreaker challenges current security measures, underscoring the critical need for more robust defenses for code completion.

AI-powered coding assistant tools have revolutionized the software engineering ecosystem. However, prior work has demonstrated that these tools are vulnerable to poisoning attacks. In a poisoning attack, an attacker intentionally injects maliciously crafted insecure code snippets into training datasets to manipulate these tools. The poisoned tools can suggest insecure code to developers, resulting in vulnerabilities in their products that attackers can exploit. However, it is still little understood whether such poisoning attacks against the tools would be practical in real-world settings and how developers address the poisoning attacks during software development. To understand the real-world impact of poisoning attacks on developers who rely on AI-powered coding assistants, we conducted two user studies: an online survey and an in-lab study. The online survey involved 238 participants, including software developers and computer science students. The survey results revealed widespread adoption of these tools among participants, primarily to enhance coding speed, eliminate repetition, and gain boilerplate code. However, the survey also found that developers may misplace trust in these tools because they overlooked the risk of poisoning attacks. The in-lab study was conducted with 30 professional developers. The developers were asked to complete three programming tasks with a representative type of AI-powered coding assistant tool, running on Visual Studio Code. The in-lab study results showed that developers using a poisoned ChatGPT-like tool were more prone to including insecure code than those using an IntelliCode-like tool or no tool. This demonstrates the strong influence of these tools on the security of generated code. Our study results highlight the need for education and improved coding practices to address new security issues introduced by AI-powered coding assistant tools.