We present CCA-Droid, a static analysis tool designed to detect cryptographic misuse related to chosen-ciphertext attacks (CCA) and chosen-plaintext attacks (CPA). CCA-Droid utilizes three key techniques: domain-specific slicing optimization to reduce analysis noise, crypto-state-aware call graph construction to capture indirect data flows via member variables, and conditional constant propagation for improved path sensitivity. Our evaluation demonstrates that CCA-Droid achieves 100% accuracy on CryptoAPI-Bench, surpassing CryptoGuard (72.3%), and maintains 94.8% accuracy on mutated code. Evaluations on the Ghera benchmark further confirm CCA-Droid's effectiveness, achieving 100% recall and 81.8% accuracy. On 16,284 real-world Android apps, CCA-Droid analyzed 96.4%, significantly outperforming existing tools such as CryptoGuard (80.4%) and QARK (40.0%). It identified cryptographic vulnerabilities in 12,678 apps (77.9%), with IV reuse, hardcoded keys, and missing authenticated encryption being the most prevalent issues.

Mobile phone numbers function as single keys to banking, government, and commerce, making the Subscriber Identity Module (SIM) a critical element of security. In April 2025, South Korea’s largest carrier experienced a SIM breach that compromised authentication keys and exposed nearly 27 million subscriber identifiers. We conducted semi-structured interviews with mental-model elicitation (N = 33) to examine user awareness, responses, and understanding of SIM-based authentication. Results reveal a pronounced awareness–action gap: participants recognized the breach yet held incomplete mental models, perceived little personal risk, and rarely acted protectively, even when affected. Learned helplessness, reliance on carriers, and the invisibility of SIM shaped these passive responses. Brief educational interventions improved conceptual understanding but seldom produced lasting behavioral change. Our findings demonstrate how technical opacity and psychological factors jointly inhibit protective action and offer design implications for usable security, emphasizing interventions that realign users’ mental models with system risks to foster sustainable practices.

Large language models (LLMs) generate human-like text, raising concerns about their misuse in creating deceptive content.Detecting LLM-generated comments (LGC) in online news is essential for preserving online discourse integrity and preventing opinion manipulation.However, effective detection faces two key challenges: the brevity and informality of news comments limit traditional detection methods, while the lack of publicly available LGC datasets hinders model development, particularly for non-English languages.To address these challenges, we propose a twofold approach.First, we develop an LGC generation framework to construct a high-quality dataset with diverse and complex examples.Second, we introduce XDAC (XAI-Driven Detection and Attribution of LLM-Generated Comments), a framework utilizing explainable AI, designed for the detection and attribution of short-form LGC in Korean news articles.XDAC leverages XAI to uncover distinguishing linguistic patterns at both token and character levels.We present the first large-scale benchmark dataset, comprising 1.3M human-written comments from Korean news platforms and 1M LLM-generated comments from 14 distinct models.XDAC outperforms existing methods, achieving a 98.5% F1 score in LGC detection with a relative improvement of 68.1%, and an 84.3% F1 score in attribution.To validate real-world applicability, we analyze 5.24M news comments from Naver, South Korea's leading online news platform, identifying 27,029 potential LLM-generated comments.

The exploration of backdoor vulnerabilities in object detectors, particularly in real-world scenarios, remains limited. A significant challenge lies in the absence of a natural physical backdoor dataset, and constructing such a dataset is both time- and labor-intensive. In this work, we address this gap by creating a large-scale dataset comprising approximately 11,800 images/frames with annotations featuring natural objects (e.g., T-shirts and hats) as triggers to incur cloaking adversarial effects in diverse real-world scenarios. This dataset is tailored for the study of physical backdoors in object detectors. Leveraging this dataset, we conduct a comprehensive evaluation of an insidious cloaking backdoor effect against object detectors, wherein the bounding box around a person vanishes when the individual is near a natural object (e.g., a commonly available T-shirt) in front of the detector. Our evaluations encompass three prevalent attack surfaces: data outsourcing, model outsourcing, and the use of pretrained models. The cloaking effect is successfully implanted in object detectors across all three attack surfaces. We extensively evaluate four popular object detection algorithms (anchor-based Yolo-V3, Yolo-V4, Faster R-CNN, and anchor-free CenterNet) using 19 videos (totaling approximately 11,800 frames) in real-world scenarios. Our results demonstrate that the backdoor attack exhibits remarkable robustness against various factors, including movement, distance, angle, non-rigid deformation, and lighting. In data and model outsourcing scenarios, the attack success rate (ASR) in most videos reaches 100% or near it, while the clean data accuracy of the backdoored model remains indistinguishable from that of the clean model, making it impossible to detect backdoor behavior through a validation set. Notably, two-stage object detectors (e.g., Faster R-CNN) show greater resistance to backdoor attacks under pure data poisoning conditions (i.e., in data outsourcing) compared to one-stage detectors (e.g., the Yolo series). However, this challenge is surmountable when the attacker controls the training process (particularly in model outsourcing), even with the same small poisoning rate budget as in data outsourcing. In transfer learning attack scenarios assessed on CenterNet, the average ASR remains high at 78%. A detailed 5-minute video illustrating our attack is available at https://youtu.be/Q3HOF4OobbY.

Large language models (LLMs) have advanced automated code generation but often produce code with critical security flaws, including buffer overflows, memory leaks, and unsafe file handling.While prior work emphasizes post-hoc vulnerability detection, we introduce a framework for secure-by-construction code generation via parameter-efficient fine-tuning (PEFT).We construct a secure training dataset by automatically fixing 7 high-impact vulnerability types in 37,540 C code samples from CodeNet, achieving 95.36% CWE reduction.We then apply prompt and prefix tuning to four open-source models (CodeGen-16B/6B-multi and StarCoder2-7B/3B), updating fewer than 1% of the parameters.On the LLMSe-cEval benchmark, our approach increases secure code generations from 20 to 36 for StarCoder2-3B and from 10 to 27 for CodeGen-6B.These results demonstrate that PEFT can substantially improve code security without full model retraining.

Linear regression is a fundamental model in data analysis, but training it over encrypted data remains expensive due to the overhead of homomorphic encryption (HE). We present PP-LR, an end-to-end CKKS-based training protocol for linear regression that supports encrypted gradient descent with feature-level parallelism and conditional bootstrapping. Compared to a standard HE implementation, it trains up to 15.7 times faster while maintaining accuracy within 0.2% of plaintext models on four real-world datasets.

File systems are essential components of modern operating systems, with Windows being one of the most dominant platforms. Recently, a series of attacks have exploited the Windows file system to trigger serious security threats such as privilege escalation. Over the past several years, dozens of such attacks have been reported and even exploited in the wild. However, Microsoft has consistently addressed these issues with targeted patches rather than fundamental redesigns — resembling a precarious game of Jenga where security measures are stacked upon an unstable foundation. In this paper, we present a five-step comprehensive analysis of the Windows file system's design weaknesses. First, we analyze how Windows differs from another operating system, Linux. Second, we investigated how these discrepancies lead to security vulnerabilities in real-world applications and identified 13 high-impact vulnerabilities, including 11 previously unknown ones. Third, we show that current compatibility layers in modern programming languages fail to handle these discrepancies properly. Specifically, we examined compatibility layers in six programming languages and found 27 non-compliant and 9 inconsistencies, rendering these layers unreliable. Fourth, through a user study involving 21 experienced developers, we found that most were unfamiliar with OS-level file system discrepancies and rarely implemented appropriate mitigations. Finally, we analyze existing countermeasures and discuss their limitations. Our findings reveal critical yet largely obscured security risks resulting from design flaws in the Windows file system. Furthermore, we suggest that Microsoft rethink its strategy and address these fundamental weaknesses.

Large language models (LLMs) have emerged as a promising tool for detecting code vulnerabilities, potentially offering advantages over traditional rule-based methods. This paper proposes an enhanced framework for vulnerability detection using LLMs, incorporating various prompt engineering strategies to improve performance. We evaluate several techniques, including role-based prompting, zero-shot chain-of-thought, and structured prompting approaches, on the DiverseVul dataset of C/C++ vulnerabilities. Our experiments assess the framework’s performance across different code structures, contextual information levels, and LLM capabilities. Our results show that using our dynamic prompt engineering technique, you can improve the F1 score by up to 100% with GPT-3.5, a widely used LLM model. We also observe that GPT-4o, Gemini 2.0 Flash, and Meta Llama 3.1 generally outperform GPT-3.5, and all models are very poor when it comes to correctly identifying the type of vulnerability in the code, with the best F1 score of 0.16 observed. However, our follow-up experiments on LLM-based vulnerability correction (i.e., patching) show a 45.77% success rate using GPT-4o, demonstrating promising results in leveraging LLMs for enhancing software security and providing insights into optimizing prompt engineering for vulnerability detection tasks.

WebAssembly (Wasm) provides a language-agnostic compilation target that delivers near-native performance for web applications, yet it also attracts adversaries who exploit Wasm to effectively steal someone else’s computer resources such as cryptojackers. While several detection tools have been proposed, their robustness against perturbations remains largely unknown.In this paper, we introduce Swamped (Systematic WebAssembly Module Perturbation Evaluation of Detectors), a framework that incorporates 22 semantics-preserving perturbation methods. Swamped generates a total of 48,840 perturbed variants from 43 cryptojacker samples and 31 additional Wasm malware binaries from real-world. We assess detection performance of six detectors: three Wasm-specific ones and three deep neural network (DNN) detectors. We find that DNN-based detectors are vulnerable to perturbations that shift the instruction distribution; profiling-based methods are disrupted by changes in instruction frequency; and semantic-aware approaches are highly sensitive to function-level dependency modifications. DNN-based detectors, which lack Wasm-specific modeling, are particularly susceptible to changes in the spatial layout of Wasm binaries. These findings highlight fundamental limitations in current Wasm malware detection approaches, relying on overly specific detection heuristics and inadequately trained or designed models. We offer suggestions to improve the robustness against perturbations.

The rapid proliferation of software vulnerabilities has created an urgent need for intelligent, automated methods to detect and mitigate security flaws at scale. Traditional vulnerability analysis depends heavily on manual inspection and domain-specific expertise, which are increasingly inadequate in the era of generative AI-driven code development. This research proposes an AI-based automated vulnerability detection and secure code generation framework that leverages multi-modal datasets, including source code and binaries, to achieve end-to- end automation across the vulnerability lifecycle: detection, patch generation, and validation. The system integrates explainable AI (XAI)-based vulnerability cause analysis, generative patch synthesis, system-level defensive code generation, Rust-based memory safety transformation, and differential privacy mechanisms for model confidentiality. Developed through a Korea- U.S. joint research initiative, this project aims to establish an internationally deployable platform for trustworthy and privacy- preserving AI -driven software security. The proposed research contributes both foundational methods and operational tools toward self-healing, explainable, and secure-by-design software ecosystems.

Large Language Models (LLMs) have rapidly advanced in reasoning capability and accessibility, driving their deployment across diverse applications. Yet this progress has also widened the surface for safety and security vulnerabilities. Adversaries can exploit prompt diversity, dialog memory, or multimodal inputs to induce unsafe or confidential outputs, while continual fine-tuning and third-party integration render static assurance infeasible. This paper introduces our ongoing national R&D project on developing the AutoPT Framework-an Autonomous Purple Teaming architecture that extends the collaborative principles of purple teaming toward self-adaptive, continuously verifiable LLM assurance. AutoPT unifies autonomous adversarial exploration and adaptive defensive reinforcement through two co-evolving agents. The red module, AutoPT-Red, employs coverage-guided fuzzing and internal measurement metrics to autonomously uncover vulnerabilities. The blue module, AutoPT-Blue, performs self-healing adaptation by updating guardrails and detecting integrity or confidentiality violations using embedding-based feedback. Preliminary case studies on jailbreak fuzzing and backdoor-poisoning defense validate the feasibility of this closed-loop, self-adapting architecture. As part of a broader national initiative, this work lays the conceptual and technical foundation for transitioning industrial purple teaming into a fully autonomous, scalable, and measurable assurance paradigm for generative AI systems.

We present a practical implicit authentication system for Virtual Reality (VR) that uses natural interaction tasksgrabbing, pointing, and typing-as behavioral biometrics. The system extracts 221 features from head-mounted and controller sensors and is trained as a lightweight SVM-based binary classifier using data from legitimate users and a small set of reference users to simulate attacker behavior. In a 24-participant study, our system achieved strong authentication performance, with median Equal Error Rates (EERs) of 0.4% for grabbing, 2.6 % for pointing, and 0.3 % for typing. Designed for on-device deployment, it requires no GPU support, completes inference within 1 second, and maintains a compact model size under 0.2 MB, enabling efficient, real-time authentication on standalone VR headsets. Security evaluations with attacker-in-the-loop experiments across no-knowledge, shoulder-surfing, and videoreplay conditions revealed clear trade-offs. Typing and pointing offered strong resistance to impersonation, while grabbing, despite high usability, was more vulnerable under video replay with a $23.8 \%$ attack success rate. These results demonstrate that secure, accurate, and real-time implicit authentication is feasible in VR, with task-specific characteristics enabling flexible deployment based on security and usability needs.

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 109guesses. 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 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.

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.

Despite stringent data protection regulations, such as the General Data Protection Regulation (GDPR), the California Consumer Privacy Act (CCPA), and other country-specific laws, numerous websites continue to use cookies to track user activities, raising significant privacy concerns. This study aims to investigate the compliance of e-commerce websites with these regulations from a cookie perspective and explore potential variations in cookie policies across different countries. We conducted a comprehensive analysis of 360 popular e-commerce websites (44,323 cookies) across multiple countries, examining cookie attributes and their potential links to privacy and security breaches. Our findings revealed that 73% of third-party cookies function as tracker cookies, with around 40% breaching lifecycle regulations. Additionally, 85% are vulnerable to potential cross-site scripting (XSS) attacks, while only 349 out of 44,323 adhere to robust measures aimed at combating cross-site request forgery (CSRF) attacks. We also discovered instances of masquerading cookies, where third-party cookies disguise themselves as first-party cookies, enabling unauthorized user tracking without consent. To the best of our knowledge, this study is the first to comprehensively analyze the compliance of e-commerce websites with the GDPR, CCPA, and country-specific regulations concerning cookie policies across different jurisdictions. Our findings highlight the urgent need for uniform and consistent cookie policies across websites and jurisdictions, as well as robust enforcement mechanisms and increased transparency to ensure compliance with data protection regulations. This research contributes to the ongoing discourse on privacy protection and underscores the importance of addressing the challenges posed by insecure cookie practices in the e-commerce sector.

Fingerprint authentication is a popular security mechanism for smartphones and laptops. However, its adoption in web and cloud environments has been limited due to privacy concerns over storing and processing biometric data on servers. This paper introduces Blind-Touch, a novel machine learning-based fingerprint authentication system leveraging homomorphic encryption to address these privacy concerns. Homomorphic encryption allows computations on encrypted data without decrypting. Thus, Blind-Touch can keep fingerprint data encrypted on the server while performing machine learning operations. Blind-Touch combines three strategies to efficiently utilize homomorphic encryption in machine learning: (1) It optimizes the feature vector for a distributed architecture, processing the first fully connected layer (FC-16) in plaintext on the client side and the subsequent layer (FC-1) post-encryption on the server, thereby minimizing encrypted computations; (2) It employs a homomorphic encryption-compatible data compression technique capable of handling 8,192 authentication results concurrently; and (3) It utilizes a clustered server architecture to simultaneously process authentication results, thereby enhancing scalability with increasing user numbers. Blind-Touch achieves high accuracy on two benchmark fingerprint datasets, with a 93.6% F1- score for the PolyU dataset and a 98.2% F1-score for the SOKOTO dataset. Moreover, Blind-Touch can match a fingerprint among 5,000 in about 0.65 seconds. With its privacy-focused design, high accuracy, and efficiency, Blind-Touch is a promising alternative to conventional fingerprint authentication for web and cloud applications.

In Android 12, Google introduced a new security feature called the privacy indicator to protect users from spyware. The privacy indicator visually alerts users by displaying a green circle in the notification bar when an application accesses the camera. While this feature initially appears effective, our work has identified two possible attack scenarios that can undermine it. The first attack uses screen overlay techniques with a higher Z-order and deceptive status bar layouts to make it difficult to see the privacy indicator. In a user study involving 44 participants, only 13.6% of participants recognized the indicator under UI overlay attacks, compared to 63.6% in default Android 12 settings. The second attack exploits device configurations to disable the privacy indicator. Our findings were reported to the developers of the Android system UI at Samsung Electronics and the Google Issue Tracker, and we received acknowledgments from both parties. As countermeasures, we recommend ensuring the integrity of the privacy indicator using trusted execution facilities. We introduce a proof-of-concept solution called SEPI (Security-Enhanced Privacy Indicator), which utilizes a secure hypervisor and ARM TrustZone. SEPI is designed to detect camera and microphone activities, subsequently displaying the relevant indicator with the highest Z-order in a securely isolated display buffer. Our experimental findings revealed only a minimal 3.3% reduction in benchmark scores compared to the device's default operational state. The SEPI privacy indicator is displayed with a negligible mean delay of 20.92 ms.

We present Blind-Match, a novel biometric identification system that leverages homomorphic encryption (HE) for efficient and privacy-preserving 1:N matching. Blind-Match introduces a HE-optimized cosine similarity computation method, where the key idea is to divide the feature vector into smaller parts for processing rather than computing the entire vector at once. By optimizing the number of these parts, Blind-Match minimizes execution time while ensuring data privacy through HE. Blind-Match achieves superior performance compared to state-of-the-art methods across various biometric datasets. On the LFW face dataset, Blind-Match attains a 99.63% Rank-1 accuracy with a 128-dimensional feature vector, demonstrating its robustness in face recognition tasks. For fingerprint identification, Blind-Match achieves a remarkable 99.55% Rank-1 accuracy on the PolyU dataset, even with a compact 16-dimensional feature vector, significantly outperforming the state-of-the-art method, Blind-Touch, which achieves only 59.17%. Furthermore, Blind-Match showcases practical efficiency in large-scale biometric identification scenarios, such as Naver Cloud's FaceSign, by processing 6,144 biometric samples in 0.74 seconds using a 128-dimensional feature vector.

AI-powered coding assistant tools (e.g., ChatGPT, Copilot, and IntelliCode) 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 (e.g., ChatGPT or IntelliCode), 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.

Conventional ad blocking and tracking prevention tools often fall short in addressing web content manipulation. Machine learning approaches have been proposed to enhance detection accuracy, yet aspects of practical deployment have frequently been overlooked. This paper introduces AdFlush, a novel machine learning model for real-world browsers. To develop AdFlush, we evaluated the effectiveness of 883 features, ultimately selecting 27 key features for optimal performance. We tested AdFlush on a dataset of 10,000 real-world websites, achieving an F1 score of 0.98, thereby outperforming AdGraph (F1 score: 0.93), WebGraph (F1 score: 0.90), and WTAgraph (F1 score: 0.84). Additionally, AdFlush significantly reduces computational overhead, requiring 56% less CPU and 80% less memory than AdGraph. We also assessed AdFlush's robustness against adversarial manipulations, demonstrating superior resilience with F1 scores ranging from 0.89 to 0.98, surpassing the performance of AdGraph and WebGraph, which recorded F1 scores between 0.81 and 0.87. A six-month longitudinal study confirmed that AdFlush maintains a high F1 score above 0.97 without the need for retraining, underscoring its effectiveness.

Establishing trust and helping experts debug and understand the inner workings of deep learning models, interpretation methods are increasingly coupled with these models, building interpretable deep learning systems. However, adversarial attacks pose a significant threat to public trust by making interpretations of deep learning models confusing and difficult to understand. In this paper, we present a novel Single-class target-specific ADVersarial attack called SingleADV. The goal of SingleADV is to generate a universal perturbation that deceives the target model into confusing a specific category of objects with a target category while ensuring highly relevant and accurate interpretations. The universal perturbation is stochastically and iteratively optimized by minimizing the adversarial loss that is designed to consider both the classifier and interpreter costs in targeted and non-targeted categories. In this optimization framework, ruled by the first- and second-moment estimations, the desired loss surface promotes high confidence and interpretation scores of adversarial samples. By avoiding unintended misclassification of samples from other categories, SingleADV enables more effective targeted attacks on interpretable deep learning systems in both white-box and black-box scenarios. To evaluate the effectiveness of SingleADV, we conduct experiments using four different model architectures (ResNet-50, VGG-16, DenseNet-169, and Inception-V3) coupled with three interpretation models (CAM, Grad, and MASK). Through extensive empirical evaluation, we demonstrate that SingleADV effectively deceives target deep learning models and their associated interpreters under various conditions and settings. Our results show that the performance of SingleADV is effective, with an average attack success rate of 74% and prediction confidence exceeding 77% on successful adversarial samples. Furthermore, we discuss several countermeasures against SingleADV, including a transfer-based learning approach and existing preprocessing defenses.

Training deep neural networks (DNNs) requires large datasets and powerful computing resources, which has led some owners to restrict redistribution without permission. Watermarking techniques that embed confidential data into DNNs have been used to protect ownership, but these can degrade model performance and are vulnerable to watermark removal attacks. Recently, DeepJudge was introduced as an alternative approach to measuring the similarity between a suspect and a victim model. While DeepJudge shows promise in addressing the shortcomings of watermarking, it primarily addresses situations where the suspect model copies the victim’s architecture. In this study, we introduce DeepTaster, a novel DNN fingerprinting technique, to address scenarios where a victim’s data is unlawfully used to build a suspect model. DeepTaster can effectively identify such DNN model theft attacks, even when the suspect model’s architecture deviates from the victim’s. To accomplish this, DeepTaster generates adversarial images with perturbations, transforms them into the Fourier frequency domain, and uses these transformed images to identify the dataset used in a suspect model. The underlying premise is that adversarial images can capture the unique characteristics of DNNs built with a specific dataset. To demonstrate the effectiveness of DeepTaster, we evaluated the effectiveness of DeepTaster by assessing its detection accuracy on three datasets (CIFAR10, MNIST, and Tiny-ImageNet) across three model architectures (ResNet18, VGG16, and DenseNet161). We conducted experiments under various attack scenarios, including transfer learning, pruning, fine-tuning, and data augmentation. Specifically, in the Multi-Architecture Attack scenario, DeepTaster was able to identify all the stolen cases across all datasets, while DeepJudge failed to detect any of the cases.

A smart contract is a self-executing program on a blockchain to ensure an immutable and transparent agreement without the involvement of intermediaries. Despite its growing popularity for many blockchain platforms like Ethereum, no technical means is available even when a smart contract requires to be protected from being copied. One promising direction to claim a software ownership is software watermarking. However, applying existing software watermarking techniques is challenging because of the unique properties of a smart contract, such as a code size constraint, non-free execution cost, and no support for dynamic allocation under a virtual machine environment. This paper introduces a novel software watermarking scheme, dubbed Smartmark, aiming to protect the ownership of a smart contract against a pirate activity. Smartmark builds the control flow graph of a target contract runtime bytecode, and locates a collection of bytes that are randomly elected for representing a watermark. We implement a full-fledged prototype for Ethereum, applying Smartmark to 27,824 unique smart contract bytecodes. Our empirical results demonstrate that Smartmark can effectively embed a watermark into a smart contract and verify its presence, meeting the requirements of credibility and imperceptibility while incurring an acceptable performance degradation. Besides, our security analysis shows that Smartmark is resilient against viable watermarking corruption attacks; e.g., a large number of dummy opcodes are needed to disable a watermark effectively, resulting in producing an illegitimate smart contract clone that is not economical.

This paper explores the usability of Docker Image Vulnerability Scanners (DIVSes) through heuristic evaluations. Docker simplifies the process of software development, distribution, deployment, and execution by providing a container-based execution environment. However, vulnerabilities in Docker images can pose security risks to containers. To mitigate this, DIVSes are crucial in helping developers identify and address these vulnerabilities in the software packages and libraries within Docker images. Despite their importance, research on the usability of DIVSes has been limited. To address this gap, we developed 11 customized heuristics and applied them to three widely-used DIVSes (Grype, Trivy, and Snyk). Our evaluations revealed 239 usability issues within the tools evaluated. Our findings highlight that the evaluated DIVSes do not provide sufficient information to comprehend the risks associated with identified vulnerabilities, prioritize them, or effectively fix them. Our study offers valuable insights and practical recommendations for enhancing the usability of DIVSes, making it easier for developers to identify and address vulnerabilities in Docker images.

Spam filtering services typically operate via cloud outsourcing, which exposes sensitive and private email content to the cloud server spam filter. Homomorphic encryption (HE) can address this issue by ensuring that user emails remain encrypted throughout all stages of the spam detection process on the cloud server. However, existing HE-based approaches are computationally infeasible due to the nature of HE operations. This paper proposes BlindFilter, a distributed, lightweight, HE-based spam email detection approach that consists of clients and servers collaborating to perform spam detection operations securely. BlindFilter employs WordPiece encoding and a modified Naive Bayes classifier, mitigating the need for multiplications and comparisons that would be prohibitive in terms of computation when applied with HE. Our experimental results demonstrate the efficacy of BlindFilter, with F1 scores exceeding 97% across two public email datasets. Furthermore, BlindFilter proves to be efficient as it can process an email in an average of 482.78 milliseconds. Our analysis also reveals that BlindFilter is robust against model extraction attacks, in which malicious users attempt to deduce the features of BlindFilter from query-response pairs.

Application fingerprinting is a useful data analysis technique for network administrators, marketing agencies, and security analysts. For example, an administrator can adopt application fingerprinting techniques to determine whether a user’s network access is allowed. Several mobile application fingerprinting techniques (e.g., FlowPrint, AppScanner, and ET-BERT) were recently introduced to identify applications using the characteristics of network traffic. However, we find that the performance of the existing mobile application fingerprinting systems significantly degrades when a virtual private network (VPN) is used. To address such a shortcoming, we propose a framework dubbed AppSniffer that uses a two-stage classification process for mobile app fingerprinting. In the first stage, we distinguish VPN traffic from normal traffic; in the second stage, we use the optimal model for each traffic type. Specifically, we propose a stacked ensemble model using Light Gradient Boosting Machine (LightGBM) and a FastAI library-based neural network model to identify applications’ traffic when a VPN is used. To show the feasibility of AppSniffer, we evaluate the detection accuracy of AppSniffer for 150 popularly used Android apps. Our experimental results show that AppSniffer effectively identifies mobile applications over VPNs with F1-scores between 84.66% and 95.49% across four different VPN protocols. In contrast, the best state-of-the-art method (i.e., AppScanner) demonstrates significantly lower F1-scores between 25.63% and 47.56% in the same settings. Overall, when normal traffic and VPN traffic are mixed, AppSniffer achieves an F1-score of 90.63%, which is significantly better than AppScanner that shows an F1-score of 70.36%.

In the age of data and machine learning, massive amounts of data produced throughout our society can be rapidly delivered to various applications through a broad spectrum of cloud services. However, the spectrum of applications has vastly different data quality requirements and Willingness-To-Pay(WTP), creating a general and complex problem matching consumer quality requirements and budgets with providers’ data quality and price. This paper proposes the Information-Centric Networking(ICN)-based data marketplace to foster quality-data trading service to address the challenge above. We embed a WTP mechanism into an ICN-based data broker service running on cloud computing; therefore, a data consumer can request its desired data with a data name and quality requirement. By specifying nominal WTPs, data consumers can acquire data of the desired quality at the range of maximum nominal WTP. At the same time, a data broker can offer data of a suitable quality based on the profit-optimized price and the proposed service quality using ground-truth accuracy trained by data. We demonstrate that the data broker’s profit can be almost doubled by using the optimal data size and budget determined by considering the one-leader-multiple-followers Stackelberg game. These results show that a value-added data brokering service can profitably facilitate data trading.

Decentralized Identity (DID) is emerging as a new digital identity management scheme that promises users complete control of their personal data and identification without central authority involvement. The World Wide Web Consortium (W3C) has drafted the DID standard and provided reference implementations. We conduct a security analysis of the W3C DID standard and the reference universal resolver implementation, focusing on user privacy in the DID resolving process. The universal resolver is the key component in the architecture that processes DID requests and DID document retrievals. Our analysis demonstrates that privacy issues can arise due to the imprudent design of the universal resolver. Furthermore, we found that side-channels in the DID document caching schemes of real-world DID services can entail privacy concerns. Motivated by our security analysis, we present a novel DID resolving design, called Oblivira, to enable obliviously DID resolving. Oblivira is a secure resolving agent with a small footprint that enforces the universal resolver to resolve requests without knowing their content. We also propose a privacy-preserving DID document caching scheme that eliminates side-channels. Our evaluation results show that Oblivira only incurs approximately 2.6% of overhead on average with different resolver settings (3, 6, and 12 threads).

The development of techniques for creating completely synthetic photographic images and videos, as well as the increasing prevalence of disinformation associated with such synthetic media, have sparked interest in the computer community. One such issue is the proliferation of deepfakes. However, much of the research is devoted to outwitting the SOTA in order to generate and detect fabricated images and video. Considering them from the perspective of computer security and human ethics is largely ignored. As such, this workshop aims to provide a forum for researchers to exchange ideas and methodologies, as well as to provide academia and industry with insights into the development and identification of fake media from a computer security perspective.

The popularity of cryptocurrency raised a new cyber security threat dubbed cryptojacking representing malicious activities for abusing victims' computing resources without their consent to mine cryptocurrency. Recently, Tekiner et al. [1] proposed an effective cryptojacking detection technique using a machine learning model with the statistical properties of the network traffic for cryptojacking in the Internet of Things (IoT) devices. In this paper, however, we demonstrate that this state-of-the-art method can effectively be evaded by maliciously manipulating the network packets for cryptojacking. Our evaluation results show that packet manipulations (packet splitting, dummy packet/payload insertion, and a proxy network) can effectively evade the model's detection -- the packet splitting technique significantly decreased the F1-score of the detection model from 0.93 to 0.30. Finally, the best combination of those packet manipulations can decrease the F1-score of the detection model to 0.21.

A user’s location information can be used to identify the user. For example, in Android, we can keep our smartphone unlocked when it is located near a place that was previously registered as a trusted place. However, existing location-based user authentication solutions failed to support fine-grained indoor location registration. In this article, we present a highly accurate identification scheme named location-based identification (LocID) using the smartphone’s physical fingerprints that are sensed at the user’s secure area. Our key idea is to use Wi-Fi signals and the light intensity of smartphones’ fingerprints to accurately identify their locations within adjacent indoor areas. To generate stable and reliable location fingerprints, we develop a novel mathematical location-wise signal signature (LSS) algorithm by minimizing the bias highly affected by various environmental factors. To show the feasibility of LocID, we evaluated the performance using machine learning classifiers with a real-world data set containing Wi-Fi and light fingerprints from 68 locations in a building for five weeks. The experimental results show that LocID achieved an F1-score of 98.3%, false-negative rate (FNR) of 2.9%, and false-positive rate (FPR) of 1.7%, respectively, when LocID was retrained daily. Moreover, LocID achieved an F1-score of 94.4%, FNR of 2.7%, and FPR of 3.7%, respectively, when LocID was trained using the data set during the first week and tested with the remaining four weeks data set. Also, we tested LocID using only one-scan of location fingerprints and achieved 90.6% and 90.8% AUC scores for support vector machine and$K$-nearest neighbor classifiers, respectively.

Federated learning (FL) and split learning (SL) are state-of-the-art distributed machine learning techniques to enable machine learning training without accessing raw data on clients or end devices. However, their comparative training performance under real-world resource-restricted Internet of Things (IoT) device settings remains barely studied. This work provides empirical comparisons of FL and SL in real-world IoT settings regarding (i) learning performance with heterogeneous data distributions and (ii) on-device execution overhead. Our analyses in this work demonstrate that the learning performance of SL is better than FL under an imbalanced data distribution but worse than FL under an extreme non-IID data distribution. Recently, FL and SL are combined to form splitfed learning (SFL) to leverage each of their benefits (e.g., parallel training of FL and lightweight on-device computation requirement of SL). Our work considers FL, SL, and SFL, and mounts them on Raspberry Pi devices to evaluate their performance, including training time, communication overhead, power consumption, and memory usage with resource-restricted IoT devices. Besides evaluations, we apply two optimizations. First, we generalize SFL by carefully examining the possibility of a hybrid type of model training at the server-side. The generalized SFL merges sequential (dependent) and parallel (independent) processes of model training and thus is beneficial to a system with a large scale of IoT devices, specifically at the server-side operations. Second, we propose pragmatic techniques to substantially reduce the communication overhead by up to four times for the SL and (generalized) SFL.

Trojan attacks on deep neural networks (DNNs) exploit abackdoorembedded in a DNN model that can hijack any input with an attacker’s chosen signature trigger. Emerging defence mechanisms are mainly designed and validated on vision domain tasks (e.g., image classification) on 2D Convolutional Neural Network (CNN) model architectures; a defence mechanism that is general across vision, text, and audio domain tasks is demanded. This work designs and evaluates a run-time Trojan detection method exploitingSTRongIntentionalPerturbation of inputs that is a multi-domain input-agnostic Trojan detection defence acrossVision,Text andAudio domains—thus termed as STRIP-ViTA. Specifically, STRIP-ViTA is demonstratively independent of not only task domain but also model architectures. Most importantly, unlike other detection mechanisms, it requires neither machine learning expertise nor expensive computational resource, which are the reason behind DNN model outsourcing scenario—one main attack surface of Trojan attack. We have extensively evaluated the performance of STRIP-ViTA over: i) CIFAR10 and GTSRB datasets using 2D CNNs for vision tasks; ii) IMDB and consumer complaint datasets using both LSTM and 1D CNNs for text tasks; and iii) speech command dataset using both 1D CNNs and 2D CNNs for audio tasks. Experimental results based on more than 30 tested Trojaned models (including publicly Trojaned model) corroborate that STRIP-ViTA performs well across all nine architectures and five datasets. Overall, STRIP-ViTA can effectively detect trigger inputs with small false acceptance rate (FAR) with an acceptable preset false rejection rate (FRR). In particular, for vision tasks, we can always achieve a 0 percent FRR and FAR given strong attack success rate always preferred by the attacker. By setting FRR to be 3 percent, average FAR of 1.1 and 3.55 percent are achieved for text and audio tasks, respectively. Moreover, we have evaluated STRIP-ViTA against a number of advanced backdoor attacks and compare its effectiveness with other recent state-of-the-arts.

To enhance the performance of web services, web servers often compress data to be delivered. Unfortunately, the data compression technique has also introduced a side effect called compression side-channel attacks (CSCA) . CSCA allows eavesdroppers to unveil secret strings included in the encrypted traffic by observing the length of data. A promising defense technique called Debreach was recently proposed to mitigate CSCA by excluding all secret data in a web page during the compression process. Although Debreach has proven to be safe against CSCA and outperforms other approaches, the exclusion of all secret data from compression eventually resulted in a decreased compression efficiency. In this paper, we present a highly efficient CSCA mitigation system called “Mutexion” ( Mut ually ex clusive compress ion ) which allows us to fully take advantage of compression over an entire web page, including secret data. The key idea behind Mutexion is to fully take advantage of all the matching subsequences within a web page except only for those between secret data and user-controlled data (potentially controlled by an attacker) during the compression process. This approach of Mutexion effectively prevents side-channel leaks of secret data under CSCA misusing user-controlled data in a web page while minimizing the degradation in compression efficiency. It is required for our compressor to trace both secret data and user-controlled data in its compression process of web pages. To meet this requirement, we provide techniques to enable automated annotation of secret and user-controlled data in web pages. We implemented Mutexion as a fully working system to test live web pages and evaluated its performance with respect to security and compression efficiency. Our evaluation results demonstrated that Mutexion effectively prevents CSCA and also achieves almost the same compression ratio as the original zlib, which is vulnerable to CSCA, with a slight increase (0.032 milliseconds (7.9%) on average) in execution time.

Recently, edge-based virtual desktop infrastructure (EdgeVDI), which brings the power of virtualized desktop infrastructure to cloudlets closer to users, has been considered as an attractive solution for WAN mobility. However, ransomware and wiper malware are becoming more and more prevalent, which can impose serious cybersecurity threats to EdgeVDI users. Existing tamper-resistant solutions cannot deal with cloudlet failures. In this paper, we propose Rocky, the first distributed replicated block device for EdgeVDI that can recover from tampering attacks and failures. The key enabler is replicating to store a consistent write sequence across cloudlets as an append-only immutable mutation history. In addition, Rocky uses a replication broker to allow heterogenous cloudlets to control replication rates at their pace and reduces both disk space and network bandwidth consumption by coalescing writes for both uplink and downlink. To show the feasibility of Rocky, we implemented Rocky in Java. The experimental results show that Rocky’s write and read throughputs are similar to those of a baseline device with 8.4% and 11.9% additional overheads, respectively. In addition, we could reduce repeated writes by 88.5% and 100% for editing presentation slides and a photo, respectively.

It is still challenging to effectively identify rumors due to rapid changes in people's interests and perceptions. To enhance rumor detectors, we first need to better understand which rumors are effective (in terms of bypassing detection) and their characteristics. In this paper, we introduce ARGH, a novel framework to automatically generate rumors using recent advancements in natural language processing, customized to target and generate specific topics. To show the effectiveness of ARGH, we conducted a user study with 212 participants and analyzed how well humans can detect the rumors generated by ARGH, and we also tested its performance against the state-of-the-art rumor detection model PLAN [17]. Surprisingly, the experimental results demonstrate that the generated rumors are significantly harder to identify as rumors than hand-written rumors, degrading the detection accuracy by both humans and machines by 18.87% and 17.62%, respectively. We believe that ARGH will be a useful tool to obtain high quality and evasive rumor datasets quickly, which is often a tedious and time consuming task. Further, our analysis results provide valuable insight into how to characterize evasive rumors and how they can be generated, which will help to enhance the existing rumor detection techniques.

Image-scaling is a typical operation that processes the input image before feeding it into convolutional neural network models. However, it is vulnerable to the newly revealed image-scaling attack. This work presents an image-scaling attack detection framework, Decamouflage, consisting of three independent detection methods: scaling, filtering, and steganalysis, to detect the attack through examining distinct image characteristics. Decamouflage has a pre-determined detection threshold that is generic. More precisely, as we have validated, the threshold determined from one dataset is also applicable to other different datasets. Extensive experiments show that Decamouflage achieves detection accuracy of 99.9% and 98.5% in the white-box and the black-box settings, respectively. We also measured its running time overhead on a PC with an Intel i5 CPU and 8GB RAM. The experimental results show that image-scaling attacks can be detected in milliseconds. Moreover, Decamouflage is highly robust against adaptive image-scaling attacks (e.g., attack image size variances).

A fundamental premise of SMS One-Time Password (OTP) is that the used pseudo-random numbers (PRNs) are uniquely unpredictable for each login session. Hence, the process of generating PRNs is the most critical step in the OTP authentication. An improper implementation of the pseudo-random number generator (PRNG) will result in predictable or even static OTP values, making them vulnerable to potential attacks. In this paper, we present a vulnerability study against PRNGs implemented for Android apps. A key challenge is that PRNGs are typically implemented on the server-side, and thus the source code is not accessible. To resolve this issue, we build an analysis tool, OTP-Lint, to assess implementations of the PRNGs in an automated manner without the source code requirement. Through reverse engineering, OTP-Lint identifies the apps using SMS OTP and triggers each app's login functionality to retrieve OTP values. It further assesses the randomness of the OTP values to identify vulnerable PRNGs. By analyzing 6,431 commercially used Android apps downloaded from Google Play and Tencent Myapp, OTP-Lint identified 399 vulnerable apps that generate predictable OTP values. Even worse, 194 vulnerable apps use the OTP authentication alone without any additional security mechanisms, leading to insecure authentication against guessing attacks and replay attacks.

A new collaborative learning, called split learning, was recently introduced, aiming to protect user data privacy without revealing raw input data to a server. It collaboratively runs a deep neural network model where the model is split into two parts, one for the client and the other for the server. Therefore, the server has no direct access to raw data processed at the client. Until now, the split learning is believed to be a promising approach to protect the client's raw data; for example, the client's data was protected in healthcare image applications using 2D convolutional neural network (CNN) models. However, it is still unclear whether the split learning can be applied to other deep learning models, in particular, 1D CNN.

Designing a fallback authentication mechanism that is both memorable and strong is a challenging problem because of the trade-off between usability and security. Security questions are popularly used as a fallback authentication method for password recovery.However, they are prone to guessing attacks by users' acquaintances and may be hard to recall. To overcome these limitations, we present PassTag, a hybrid password scheme that takes advantage of both graphical and textual password authentication methods. PassTag combines a user-provided image and a short personalized text description of the image, imagetag, as an authentication secret.Furthermore, PassTag incorporates decoy images to make it difficult to guess the user-provided pictures. We conducted three user studies with 161 participants for up to three months to evaluate the performance of PassTag against security questions. The evaluation results demonstrate that PassTag is significantly stronger against close adversaries and highly memorable (92.6%-95.0%) after one,two, and three months, respectively. Our longitudinal study results show PassTag is a promising alternative for fallback authentication.

Touchscreen gestures are attracting research attention as an authentication method. While studies have showcased their usability, it has proven more complex to determine, let alone enhance, their security. Problems stem both from the small scale of current data sets and the fact that gestures are matched imprecisely - by a distance metric. This makes it challenging to assess entropy with traditional algorithms. To address these problems, we captured a large set of gesture passwords (N=2594) from crowd workers, and developed a security assessment framework that can calculate partial guessing entropy estimates, and generate dictionaries that crack 23.13% or more gestures in online attacks (within 20 guesses). To improve the entropy of gesture passwords, we designed novel blacklist and lexical policies to, respectively, restrict and inspire gesture creation. We close by validating both our security assessment framework and policies in a new crowd-sourced study (N=4000). Our blacklists increase entropy and resistance to dictionary based guessing attacks.

Federated learning (FL) and split neural networks (SplitNN) are state-of-art distributed machine learning techniques to enable machine learning without directly accessing raw data on clients or end devices. In theory, such distributed machine learning techniques have great potential in distributed applications, in which data are typically generated and collected at the client-side while the collected data should be processed by the application deployed at the server-side. However, there is still a significant gap in evaluating the performance of those techniques concerning their practicality in the Internet of Things (IoT)-enabled distributed systems constituted by resource-constrained devices. This work is the first attempt to provide empirical comparisons of FL and SplitNN in real-world IoT settings in terms of learning performance and device implementation overhead. We consider a variety of datasets, different model architectures, multiple clients, and various performance metrics. For the learning performance (i.e., model accuracy and convergence time), we empirically evaluate both FL and SplitNN under different types of data distributions such as imbalanced and non-independent and identically distributed (non-IID) data. We show that the learning performance of SplitNN is better than FL under an imbalanced data distribution but worse than FL under an extreme non-IID data distribution. For implementation overhead, we mount both FL and SplitNN on Raspberry Pi devices and comprehensively evaluate their overhead, including training time, communication overhead, power consumption, and memory usage. Our key observations are that under the IoT scenario where the communication traffic is the primary concern, FL appears to perform better over SplitNN because FL has a significantly lower communication overhead compared with SplitNN. However, our experimental results also demonstrate that neither FL or SplitNN can be applied to a heavy model, e.g., with several million parameters, on resource-constrained IoT devices because its training cost would be too expensive for such devices. Source code is released and available: https://github.com/Minki-Kim95/Federated-Learning-and-Split-Learning-with-raspberry-pi.

A package tracking number (PTN) is widely used to monitor and track a shipment. Through the lenses of security and privacy, however, a package tracking number can possibly reveal certain personal information, leading to security and privacy breaches. In this work, we examine the privacy issues associated with online package tracking systems used in the top three most popular package delivery service providers (FedEx, DHL, and UPS) in the world and found that those websites inadvertently leak users’ personal data with a PTN. Moreover, we discovered that PTNs are highly structured and predictable. Therefore, customers’ personal data can be massively collected via PTN enumeration attacks. We analyzed more than one million package tracking records obtained from Fedex, DHL, and UPS, and showed that within 5 attempts, an attacker can efficiently guess more than 90% of PTNs for FedEx and DHL, and close to 50% of PTNs for UPS. In addition, we present two practical attack scenarios: 1) to infer business transactions information and 2) to uniquely identify recipients. Also, we found that more than 109 recipients can be uniquely identified with less than 10 comparisons by linking the PTN information with the online people search service, Whitepages.

To accelerate the deployment of fifth-generation (5G) cellular networks, millions of devices are being connected to massive Internet of Things (IoT) networks. However, advances in the scale of connectivity on 5G networks may increase the attack surface of these devices, thereby increasing the number of attack opportunities. To address the potential security risks in IoT systems, one feasible security practice involves the development of secure and efficient user authentication schemes. In 2017, Dhillon and Kalra proposed a three-factor user authentication scheme for IoT. We noted that their scheme suffers from several security weaknesses. In this study, we specifically demonstrate that the scheme proposed by Dhillon and Kalra (1) is not secured from a stolen mobile device attack; (2) does not prevent a user impersonation attack; (3) does not provide a session key agreement; (4) does not have a contingency plan (e.g., a revocation phase) for situations where a user’s private key is compromised, or a mobile device is stolen or lost. We propose an improved three-factor user authentication scheme to resolve these security issues. Furthermore, we demonstrate that the proposed scheme provides desirable attributes for IoT environments and that its computation and communication costs are suitable for extremely low-cost IoT devices.

To help smartphone users protect their phone, fingerprint-based authentication systems (e.g., Apple’s Touch ID) have increasingly become popular in smartphones. In web applications, however, fingerprint-based authentication is still rarely used. One of the most serious concerns is the lack of technology for securely storing fingerprint data used for authentication. Because scanned fingerprint data are not exactly the same each time, the use of a traditional cryptographic hash function (e.g., SHA-256) is infeasible to protect raw fingerprint data. In this paper, we present an efficient privacy-preserving fingerprint authentication system using a fully homomorphic encryption scheme in which fingerprint data are always stored and processed in an encrypted form. We implement a fully working fingerprint authentication system with a fingerprint database (containing 4,000 samples) using the Fast Fully Homomorphic Encryption over the Torus (TFHE) library. The proposed system can perform the fingerprint matching process within about 166 seconds (±0.564 seconds) on average.

The latest smartphones have started providing multiple authentication options including PINs, patterns, and passwords (knowledge based), as well as face, fingerprint, iris, and voice identification (biometric-based). In this article, we conducted two user studies to investigate how the convenience and security of unlocking phones are influenced by the provision of multiple authentication options. In a task-based user study with 52 participants, we analyze how participants choose an option to unlock their smartphone in daily life. The user study results demonstrate that providing multiple biometric-based authentication choices does not really influence convenience, because fingerprint had monopolistic dominance in the usage of unlock methods (111 of a total of 115 unlock trials that used a biometric-based authentication factor) due to users’ habitual behavior and fastness in unlocking phones. However, convenience was influenced by the provision of both knowledge-based and biometric-based authentication categories, as biometric-based authentication options were used in combination with knowledge-based authentication options—pattern was another frequently used unlock method. Our findings were confirmed and generalized through a follow-up survey with 327 participants. First, knowledge-based and biometric-based authentication options are used interchangeably. Second, providing multiple authentication options for knowledge-based authentication may influence convenience—both PINs (55.7%) and patterns (39.2%) are quite evenly used. Last, in contrast to knowledge-based authentication, providing multiple authentication choices for biometric-based authentication has less influence on choosing unlock options—fingerprint scanner is the most frequently used option (134 of 187 unlock methods used among biometric-based authentication options).

Currently, miners typically join mining pools to solve cryptographic puzzles together, and mining pools are in high competition. This has led to the development of several attack strategies such as block withholding (BWH) and fork after withholding (FAW) attacks that can weaken the health of PoW systems and but maximize mining pools' profits. In this paper, we present strategies called Adaptive Retaliation Strategies (ARS) to mitigate not only BWH attacks but also FAW attacks. In ARS, each pool cooperates with other pools in the normal situation, and adaptively executes either FAW or BWH attacks for the purpose of retaliation only when attacked. In addition, in order for rational pools to adopt ARS, ARS should strike to an adaptive balance between retaliation and selfishness because the pools consider their payoff even when they retaliate. We theoretically and numerically show that ARS would not only lead to the induction of a no-attack state among mining pools, but also achieve the adaptive balance between retaliation and selfishness.

Many companies offer automatic speech recognition or Speech-to-Text APIs for use in diverse applications. However, audio classification algorithms trained with deep neural networks (DNNs) can sometimes misclassify adversarial examples, posing a significant threat to critical applications. In this paper, we present a novel way to create adversarial audio examples using a genetic algorithm. Our algorithm creates adversarial examples by iteratively adding perturbations to the original audio signal. Unlike most white-box adversarial example generations, our approach does not require knowledge about the target DNN's model and parameters (black-box) and heavy computational power of GPU resources. To show the feasibility of the proposed idea, we implement a tool called, Foolgle, using a genetic algorithm that performs untargeted attacks to create adversarial audio examples and evaluate those with the state-of-the-art Google Cloud Speech-to-Text API. Our preliminary experiment results show that Foolgle deceives the API with a success probability of 86%.

Decompilation is frequently used to analyze binary programs. In Android, however, decompilers all perform differently with varying apps due to their own characteristics. Obviously, there is no universal solution in all conditions. Based on this observation, we present a practical Android app decompilation system (called Kerberoid) that automatically stitches the results from multiple decompilers together to maximize the coverage and the accuracy of decompiled codes. We evaluate the performance of Kerberoid with 151 Android apps in which their corresponding source codes are publicly available. Kerberoid fully recovered all functions for 17% of the apps tested and gained a similarity score over 50% for 40% of the apps tested, increased by 7% and 9%, respectively, compared with the best existing decompiler.

Sharding-based consensus protocols were introduced to enable the parallelization of the consensus work and storage for blockchain systems. However, existing sharding-based consensus algorithms are not sufficiently designed for distributing miners and transactions to shards in a fair and secure manner, which would make the blockchain systems vulnerable to several attacks. To overcome such limitations of the existing sharding-based consensus protocols, we present a new sharding-based Proof-of-Stake (PoS) blockchain protocol using fair and dynamic sharding management. To show the security of the proposed consensus protocol, we numerically analyze attack probabilities and found that the proposed protocol is secure when the number of shards is less than or equal to 6. Moreover, the proposed protocol is approximately 186 times more efficient than Ethereum with the real parameter settings obtained by the Ethereum network.

To understand users' risk perceptions about sharing their PHR on SNS, we first conducted a qualitative user study by interviewing 16 participants. Next, we conducted a large-scale online user study with 497 participants in the U.S. to validate our qualitative results from the first study. Our study results show that a majority of users do not show strong motivation of sharing PHR on SNS due to several concerns such as misuse/abuse and security issues of shared PHR. In particular, participants are highly concerned about sharing diseases and diagnostic test results than other types of PHRs (e.g., details of hospital visits and medical interviews). However, we found about 55.13% of the participants have the experiences of sharing their health-related information on SNS. Also, we learned that users' sharing behavior for disease data can be significantly influenced by the severity/type of the disease as well as preferred recipients. Based on our findings, we propose a recommendation method to automatically determine whether users' posts can be shared with everyone on SNS by analyzing the keywords frequently occurred in health-related posts. Our implementation using Random Forest achieved an F-measure of 87.4%, indicating that PHR can be restrictively shared with high accuracy, when sharing health-related posts on SNS.

Existing continuous authentication proposals tend to have two major drawbacks. First, touch-based smartphone authentication approaches typically require explicit user interactions with the smartphone to collect sufficient touch data. These approaches may provide an attacker the opportunity to steal a victim's sensitive data before the system detects the attacker's intrusion. Likewise, an attacker may disable the continuous authentication scheme itself before detection. Second, sensor-based continuous authentication approaches inherently suffer from high energy consumption due to the constant usage of multiple sensors. In this paper, we present a novel continuous authentication system that collects light sensor data from a user's smartphone and analyzes them to authenticate users using support vector machines. We focus on the possibility of collecting light sensor data from users' smartphones while they are conducting daily behaviors to develop an anomaly detection system.

Recently, Samsung released Galaxy S10 supporting the cryptowallet feature [5]. However, it is still questionable whether cryptocurrency can be popularly used for mobile payments because processing transactions in existing blockchain systems are too slow. For example, public blockchain systems such as Bitcoin [6] (7 transactions per second (TPS)) and Ethereum (15 TPS) are significantly slower than mainstream payment systems such as Visa (2,000 TPS) using a centralized database.

There are advances in detecting malware using machine learning (ML), but it is still a challenging task to detect advanced malware variants (e.g., polymorphic and metamorphic variations). To detect such variants, we first need to understand the methods used to generate them to bypass the detection methods. In this paper, we introduce an adaptive malware variant generation (AMVG) framework to study bypassing malware detection methods efficiently. The AMVG framework uses ML (e.g., genetic algorithm (GA)) to generate malware variants that satisfy specific detection criteria. The use of GA automates the malware variant generations with appropriate modules to handle various input formats. For the experiment, we use malware samples retrieved from theZoo, a collection of malware samples. The results show that we can automatically generate malware variants that satisfy varying detection criteria in a practical amount of time, as well as showing the capabilities to handle different input formats.

The Digital economy is based on confidence in its trustworthiness. Blockchain distributed consensus provides a reliable and trustful network for financial and non-financial transactions. Blockchain-based electric vehicles (EVs) charging applications benefit blockchain features to provide automated and verifiable services for EV charging market. Requirements for feasible charging operation and privacy concerns are challenging issues with blockchain-based EV charging approaches. To provide a feasible charging ability and preserve EV owner's privacy, we introduce EVChain. The EVChain is a trustful and decentralized platform based on blockchain technology to share charging credits in the EV charging market. To share credits, the main blockchain in EVChain is connected to one or more subnetwork blockchains. We introduce an interconnection position to preserve EV owners' privacy with k-anonymity protection. We simulate and evaluate the privacy protection it provides, based on an example EV charging scenario.

Bitcoin has become the most popular cryptocurrency based on a peer-to-peer network. In Aug. 2017, Bitcoin was split into the original Bitcoin (BTC) and Bitcoin Cash (BCH). Since then, miners have had a choice between BTC and BCH mining because they have compatible proof-of-work algorithms. Therefore, they can freely choose which coin to mine for higher profit, where the profitability depends on both the coin price and mining difficulty. Some miners can immediately switch the coin to mine only when mining difficulty changes because the difficulty changes are more predictable than that for the coin price, and we call this behavior fickle mining. In this paper, we study the effects of fickle mining by modeling a game between two coins. To do this, we consider both fickle miners and some factions (e.g., BITMAIN for BCH mining) that stick to mining one coin to maintain that chain. In this model, we show that fickle mining leads to a Nash equilibrium in which only a faction sticking to its coin mining remains as a loyal miner to the less valued coin (e.g., BCH), where loyal miners refer to those who conduct mining even after coin mining difficulty increases. This situation would cause severe centralization, weakening the security of the coin system. To determine which equilibrium the competing coin systems (e.g., BTC vs. BCH) are moving toward, we traced the historical changes of mining power for BTC and BCH and found that BCH often lacked loyal miners until Nov. 13, 2017, when the difficulty adjustment algorithm of BCH mining was changed. However, the change in difficulty adjustment algorithm of BCH mining led to a state close to the stable coexistence of BTC and BCH. We also demonstrate that the lack of BCH loyal miners may still be reached when a fraction of miners automatically and repeatedly switches to the most profitable coin to mine (i.e., automatic mining). According to our analysis, as of Dec. 2018, loyal miners to BCH would leave if more than about 5% of the total mining capacity for BTC and BCH has engaged in the automatic mining. In addition, we analyze the recent “hash war” between Bitcoin ABC and SV, which confirms our theoretical analysis. Finally, we note that our results can be applied to any competing cryptocurrency systems in which the same hardware (e.g., ASICs or GPUs) can be used for mining. Therefore, our study brings new and important angles in competitive coin markets: a coin can intentionally weaken the security and decentralization level of the other rival coin when mining hardware is shared between them, allowing for automatic mining.

Because of the dynamic nature of network traffic patterns, such as new traffic application arrivals or flash events, it is becoming increasingly difficult for conventional anomaly detection systems to separate various applications based on their traffic patterns. In this paper, by leveraging transport layer packet-level and flow-level features, new structures called application fingerprints are generated, which express such features in a compact and efficient manner. Based on the generated fingerprints, we propose a novel traffic classification framework. The proposed system generates profiles of normal applications using a multi-modal probability distribution. The proposed classification framework is then extended to detect distributed denial of service attacks from the collected statistical information at flow level. To demonstrate the feasibility of the proposed system, we evaluate its performance using five real-world traffic datasets. The experiment results show that the proposed method is capable of achieving an accuracy of over 97%, whereas the misclassification rate is only 2.5%.

For long-distance flying, drones often need to charge their battery at quick battery-charging machines (QCMs) because of their limited battery capacity. If a drone individually chooses a QCM without any coordination, a drone network may experience QCM congestion when multiple drones select the same QCM. This QCM congestion may lead to an increasing drone traffic delay. In order to solve this problem, we propose a cloud-based drone navigation (CBDN) system for efficient drone battery charging in drone networks. In order to achieve this goal, the CBDN gathers drone traffic information and determines efficient drone routes so as to minimize the overall QCM congestion level for drone battery charging using cloud-based management. Our key idea is to find globally coordinated drone routes so as to minimize the total traffic delay in a drone network by reducing the overall QCM congestion level. In order to demonstrate the effectiveness of the proposed system, we evaluated the performance of CBDN by simulating a drone network under various network conditions. The simulation results show that CBDN is more efficient than the existing shortest-path-based drone route planning algorithms in terms of end-to-end traffic delay and QCM average utilization.

Mining for crypto currencies is usually performed on high-performance single purpose hardware or GPUs. However, mining can be easily parallelized and distributed over many less powerful systems. Cryptojacking is a new threat on the Internet and describes code included in websites that uses a visitor's CPU to mine for crypto currencies without the their consent. This paper introduces MiningHunter, a novel web crawling framework which is able to detect mining scripts even if they obfuscate their malicious activities. We scanned the Alexa Top 1 million websites for cryptojacking, collected more than 13,400,000 unique JavaScript files with a total size of 246 GB and found that 3,178 websites perform cryptocurrency mining without their visitors' consent. Furthermore, MiningHunter can be used to provide an in-depth analysis of cryptojacking campaigns. To show the feasibility of the proposed framework, three of such campaigns are examined in detail. Our results provide the most comprehensive analysis to date of the spread of cryptojacking on the Internet.

A Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) is a defensive mechanism designed to differentiate humans and computers to prevent unauthorized use of online services by automated attacks. They often consist of a visual or audio test that humans can perform easily but that bots cannot solve. However, with current machine learning techniques and open-source neural network architectures, it is now possible to create a self-contained system that is able to solve specific CAPTCHA types and outperform some human users. In this paper, we present a neural network that leverages Mozilla's open source implementation of Baidu's Deep Speech architecture; our model is currently able to solve the audio version of an open-source CATPCHA system (named SimpleCaptcha) with 98.8% accuracy. Our network was trained on 100,000 audio samples generated from SimpleCaptcha and can solve new SimpleCaptcha audio tests in 1.25 seconds on average (with a standard deviation of 0.065 seconds). Our implementation seems additionally promising because it does not require a powerful server to function and is robust to adversarial examples that target Deep Speech's pre-trained models.

A CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) provides the first line of defense to protect websites against bots and automatic crawling. Recently, audio-based CAPTCHA systems are started to use for visually impaired people in many internet services. However, with the recent improvement of speech recognition and machine learning system, audio CAPTCHAs have come to struggle to distinguish machines from users, and this situation will likely continue to worsen. Unlike conventional CAPTCHA systems, we propose a new conceptual audio CAPTCHA system, combining certain sound, which is only understandable by a machine. Our experiment results demonstrate that the tested speech recognition systems always provide correct responses for our CAPTCHA samples while humans cannot possibly understand them. Based on this computational gap between the human and machine, we can detect bots with their correct responses, rather than their incorrect ones.

Even though biometric authentication such as fingerprint authentication is popularly used, there are few network services supporting biometric authentication because many users have serious privacy concerns about leakage of the biometric data on a server. For example, in fingerprint authentication, a user's raw fingerprint is typically stored in plaintext form. Unlike conventional text passwords, we cannot use a cryptographic hash function such as SHA-256 because it is very hard to obtain the same fingerprint features even when the exactly same finger is scanned multiple times. In this paper, we present a fingerprint authentication mechanism using the Fastest Homomorphic Encryption in the West (FHEW) library for Learning With Errors (LWE) scheme. Our implementation allows us to securely store fingerprint data on a network server using homomorphic encryption to calculate the distance between two encrypted fingerprint data without decrypting them. To show the efficiency of our implementation, we used "NIST special database 9'' containing 4,000 fingerprint samples. Our results show that two fingerprints can be efficiently compared with about 209 seconds on average even when they are securely stored in encrypted form.

Smartwatches support access to a wide range of private information but little is known about the security and usability of existing smartwatch screen lock mechanisms. Prior studies suggest that smartwatch authentication via standard techniques such as 4-digit PINs is challenging and error-prone. We conducted interviews to shed light on current practices, revealing that smartwatch users consider the ten-key keypad required for PIN entry to be hard to use due to its small button sizes. To address this issue, we propose the Personal Identification Chord (PIC), an authentication system based on a four-button chorded keypad that enables users to enter ten different inputs via taps to one or two larger buttons. Two studies assessing usability and security of our technique indicate PICs lead to increases in setup and (modestly) recall time, but can be entered accurately while maintaining high recall rates and may improve guessing entropy compared to PINs.

This paper presents a possible implementation of progressive authentication using the Android pattern lock. Our key idea is to use one pattern for two access levels to the device; an abridged pattern is used to access generic applications and a second, extended and higher-complexity pattern is used less frequently to access more sensitive applications. We conducted a user study of 89 participants and a consecutive user survey on those participants to investigate the usability of such a pattern scheme. Data from our prototype showed that for unlocking lowsecurity applications the median unlock times for users of the multiple pattern scheme and conventional pattern scheme were 2824 ms and 5589 ms respectively, and the distributions in the two groups differed significantly (Mann-Whitney U test, p-value less than 0.05, two-tailed). From our user survey, we did not find statistically significant differences between the two groups for their qualitative responses regarding usability and security (t-test, p-value greater than 0.05, two-tailed), but the groups did not differ by more than one satisfaction rating at 90% confidence.

Electrical network frequency (ENF) signals have common patterns that can be used as signatures for identifying recorded time and location of videos and sound. To enable cost-efficient, reliable and scalable location inference, we created a reference map of ENF signals representing hundreds of locations world wide -- extracting real-world ENF signals from online multimedia streaming services (e.g., YouTube and Explore). Based on this reference map of ENF signals, we propose a novel side-channel attack that can identify the physical location of where a target video or sound was recorded or streamed from. Our attack does not require any expensive ENF signal receiver nor any software to be installed on a victim»s device -- all we need is the recorded video or sound files to perform the attack and they are collected from world wide web. The evaluation results show that our attack can infer the intra-grid location of the recorded audio files with an accuracy of $76$% when those files are $5$ minutes or longer. We also showed that our proposed attack works well even when video and audio data are processed within a certain distortion range with audio codecs used in real VoIP applications.

Online game involves a very large number of users who are interconnected and interact with each other via the Internet. We studied the characteristics of exchanging virtual goods with real money through the processes called "real money trading (RMT)". This exchange might influence online game user behaviors and cause damage to the reputation of game companies. We examined in-game transactions to reveal RMT by constructing a social graph of virtual goods exchanges in an online game and identifying network communities of users.

Establishing secret command and control (C&C) channels from attackers is important in malware design. This paper presents design and analysis of malware architecture exploiting push notification services as C&C channels. The key feature of the push notification-based malware design is remote triggering , which allows attackers to trigger and execute their malware by push notifications. The use of push notification services as covert channels makes it difficult to distinguish this type of malware from other normal applications also using the same services. We implemented a backdoor prototype on Android devices as a proof-of-concept of the push notification-based malware and evaluated its stealthiness and feasibility. Our malware implementation effectively evaded the existing malware analysis tools such as 55 antimalware scanners from VirusTotal and SandDroid. In addition, our backdoor implementation successfully cracked about 98% of all the tested unlock secrets (either PINs or unlock patterns) in 5 seconds with only a fraction (less than 0.01%) of the total power consumption of the device. Finally, we proposed several defense strategies to mitigate push notification-based malware by carefully analyzing its attack process. Our defense strategies include filtering subscription requests for push notifications from suspicious applications, providing centralized management and access control of registration tokens of applications, detecting malicious push messages by analyzing message contents and characteristic patterns demonstrated by malicious push messages, and detecting malware by analyzing the behaviors of applications after receiving push messages.

At present, a number of users employ an authentication protocol so as to enjoy protected electronic transactions in wireless networks. In order to establish an efficient and robust the transaction system, numerous researches have been conducted relating to authentication protocols. Recently, Kaul and Awasthi presented an user authentication and key agreement scheme, arguing that their scheme is able to resist various types of attacks and preserve diverse security properties. However, this scheme possesses critical vulnerabilities. First, the scheme cannot prevent two kinds of attacks, including off-line password guessing attacks and user impersonation attacks. Second, user anonymity rule cannot be upheld. Third, session key can be compromised by an attacker. Fourth, there is high possibility that the time synchronization trouble occurs. Therefore, we suggest an upgraded version of the user authenticated key agreement method that provides enhanced security. Our security and performance analysis shows that compared, to other associated protocols, our method not only improves the security level but also ensures efficiency.

Content‐Centric Networking (CCN) is considered as a promising alternative to traditional IP‐based networking for vehicle‐to‐everything communication environments. In general, CCN packets must be fragmented and reassembled based on the Maximum Transmission Unit (MTU) size of the content delivery path. It is thus challenging to securely protect fragmented packets against attackers who intentionally inject malicious fragments to disrupt normal services on CCN‐based vehicular networks. This paper presents a new secure content fragmentation method that is resistant to Denial‐of‐Service (DoS) attacks in CCN‐based vehicular networks. Our approach guarantees the authenticity of each fragment through the immediate fragment verification at interim nodes on the routing path. Our experiment results demonstrate that the proposed approach provides much stronger security than the existing approach named FIGOA, without imposing a significant overhead in the process. The proposed method achieves a high immediate verification probability of 98.2% on average, which is 52% higher than that of FIGOA, while requiring only 14% more fragments than FIGOA.

Many machine learning methods make assumptions about the data, such as, data stationarity and data independence, for an efficient learning process that requires less data. However, these assumptions may give rise to vulnerabilities if violated by smart adversaries. In this paper, we propose a novel algorithm to craft the input samples by modifying a certain fraction of input features as small as in order to bypass the decision boundary of widely used binary classifiers using Support Vector Machine (SVM). We show that our algorithm can reliably produce adversarial samples which are misclassified with 98% success rate while modifying 22% of the input features on average. Our goal is to evaluate the robustness of classification algorithms for high demensional network data by intentionally performing evasion attacks with carefully designed adversarial examples. The proposed algorithm is evaluated using real network traffic datasets (CAIDA 2007 and CAIDA 2016).

Android allows 20 consecutive fail attempts on unlocking a device. This makes it difficult for pure guessing attacks to crack user patterns on a stolen device before it permanently locks itself. We investigate the effectiveness of combining Markov model-based guessing attacks with smudge attacks on unlocking Android devices within 20 attempts. Detected smudges are used to pre-compute all the possible segments and patterns, significantly reducing the pattern space that needs to be brute-forced. Our Markov-model was trained using 70% of a real-world pattern dataset that consists of 312 patterns. We recruited 12 participants to draw the remaining 30% on Samsung Galaxy S4, and used smudges they left behind to analyze the performance of the combined attack. Our results show that this combined method can significantly improve the performance of pure guessing attacks, cracking 74.17% of patterns compared to just 13.33% when the Markov model-based guessing attack was performed alone---those results were collected from a naive usage scenario where the participants were merely asked to unlock a given device. Even under a more complex scenario that asked the participants to use the Facebook app for a few minutes---obscuring smudges were added as a result---our combined attack, at 31.94%, still outperformed the pure guessing attack at 13.33%. Obscuring smudges can significantly affect the performance of smudge-based attacks. Based on this finding, we recommend that a mitigation technique should be designed to help users add obscurity, e.g., by asking users to draw a second random pattern upon unlocking a device.

A common security practice used to deal with a password breach is locking user accounts and sending out an email to tell users that they need to reset their password to unlock their account. This paper evaluates the effectiveness of this security practice based on the password reset email that LinkedIn sent out around May 2016, and through an online survey conducted on 249 LinkedIn users who received that email. Our evaluation shows that only about 46% of the participants reset their passwords. The mean time taken to reset password was 26.3 days, revealing that a significant proportion of the participants reset their password a few weeks, or even months after first receiving the email. Our findings suggest that more effective persuasive measures need to be added to convince users to reset their password in a timely manner, and further reduce the risks associated with delaying password resets.

To improve the security of user-chosen Android screen lock patterns, we propose a novel system-guided pattern lock scheme called "SysPal" that mandates the use of a small number of randomly selected points while selecting a pattern. Users are given the freedom to use those mandated points at any position. We conducted a large-scale online study with 1,717 participants to evaluate the security and usability of three SysPal policies, varying the number of mandatory points that must be used (upon selecting a pattern) from one to three. Our results suggest that the two SysPal policies that mandate the use of one and two points can help users select significantly more secure patterns compared to the current Android policy: 22.58% and 23.19% fewer patterns were cracked. Those two SysPal policies, however, did not show any statistically significant inferiority in pattern recall success rate (the percentage of participants who correctly recalled their pattern after 24 hours). In our lab study, we asked participants to install our screen unlock application on their own Android device, and observed their real-life phone unlock behaviors for a day. Again, our lab study did not show any statistically significant difference in memorability for those two SysPal policies compared to the current Android policy.

This paper presents a new type of online password guessing attack called "WiPING" (Wi-Fi signal-based PIN Guessing attack) to guess a victim's PIN (Personal Identification Number) within a small number of unlock attempts. WiPING uses wireless signal patterns identified from observing sequential finger movements involved in typing a PIN to unlock a mobile device. A list of possible PIN candidates is generated from the wireless signal patterns, and is used to improve performance of PIN guessing attacks. We implemented a proof-of-concept attack to demonstrate the feasibility of WiPING. Our results showed that WiPING could be practically effective: while pure guessing attacks failed to guess all 20 PINs, WiPING successfully guessed two PINs.

Smartphone advertisement is increasingly used among many applications and allows developers to obtain revenue through in-app advertising. Our study aims at identifying potential security risks of mobile-based advertising services where advertisers are charged for their advertisements on mobile applications. In the Android platform, we particularly implement bot programs that can massively generate click events on advertisements on mobile applications and test their feasibility with eight popular advertising networks. Our experimental results show that six advertising networks (75 %) out of eight are vulnerable to our attacks. To mitigate click fraud attacks, we suggest three possible defense mechanisms: (1) filtering out program-generated touch events; (2) identifying click fraud attacks with faked advertisement banners; and (3) detecting anomalous behaviors generated by click fraud attacks. We also discuss why few companies were only willing to deploy such defense mechanisms by examining economic misincentives on the mobile advertising industry.

PassBYOP is a new graphical password scheme for public terminals that replaces the static digital images typically used in graphical password systems with personalized physical tokens, herein in the form of digital pictures displayed on a physical user-owned device such as a mobile phone. Users present these images to a system camera and then enter their password as a sequence of selections on live video of the token. Highly distinctive optical features are extracted from these selections and used as the password. We present three feasibility studies of PassBYOP examining its reliability, usability, and security against observation. The reliability study shows that image-feature based passwords are viable and suggests appropriate system thresholds - password items should contain a minimum of seven features, 40% of which must geometrically match originals stored on an authentication server in order to be judged equivalent. The usability study measures task completion times and error rates, revealing these to be 7.5 s and 9%, broadly comparable with prior graphical password systems that use static digital images. Finally, the security study highlights PassBYOP's resistance to observation attack - three attackers are unable to compromise a password using shoulder surfing, camera-based observation, or malware. These results indicate that PassBYOP shows promise for security while maintaining the usability of current graphical password schemes.

Smartphone advertisement is increasingly used among many applications and allows developers to obtain revenue through in-app advertising. Our study aims at identifying potential security risks of a type of mobile advertisement where advertisers are charged for their advertisements only when a user clicks (or touches) on the advertisements in their applications. In the Android platform, we design an automated click generation attack and empirically evaluate eight popular advertising networks by performing real attacks on them. Our experimental results show that six advertising networks (75%) out of eight (Millennial Media, App Lovin, Ad Fit, Mdot M, Rev Mob and Cauly Ads) are vulnerable to our attacks. We also discuss how to develop effective defense mechanisms to mitigate such automated click fraud attacks.

System-generated random passwords have maximum password security and are highly resistant to guessing attacks. However, few systems use such passwords because they are difficult to remember. In this paper, we propose a system-initiated password scheme called "Surpass" that lets users replace few characters in a random password to make it more memorable. We conducted a large-scale online study to evaluate the usability and security of four Surpass policies, varying the number of character replacements allowed from 1 to 4 in randomly-generated 8-character passwords. The study results suggest that some Surpass policies (with 3 and 4 character replacements) outperform by 11% to 13% the original randomly-generated password policy in memorability, while showing a small increase in the percentage of cracked passwords. When compared to a user-generated password complexity policy (that mandates the use of numbers, symbols, and uppercase letters) the Surpass policy with 4-character replacements did not show statistically significant inferiority in memorability. Our qualitative lab study showed similar trends. This Surpass policy demonstrated significant superiority in security though, with 21% fewer cracked passwords than the user-generated password policy.

We propose an effective pattern lock strength meter to help users choose stronger pattern locks on Android devices. To evaluate the effectiveness of the proposed meter with a real world dataset (i.e., with complete ecological validity), we created an Android application called EnCloud that allows users to encrypt their Dropbox files. 101 pattern locks generated by real EnCloud users were collected and analyzed, where some portion of the users were provided with the meter support. Our statistical analysis indicates that about 10% of the pattern locks that were generated without the meter support could be compromised through just 16 guessing attempts. As for the pattern locks that were generated with the meter support, that number goes up to 48 guessing attempts, showing significant improvement in security. Our recommendation is to implement a strength meter in the next version of Android.

Honey encryption (HE) is a new technique to overcome the weakness of conventional password-based encryption (PBE). However, conventional honey encryption still has the limitation that it works only for binary bit streams or integer sequences because it uses a fixed distribution-transforming encoder (DTE). In this paper, we propose a variant of honey encryption called visual honey encryption which employs an adaptive DTE in a Bayesian framework so that the proposed approach can be applied to more complex domains including images and videos. We applied this method to create a new steganography scheme which significantly improves the security level of traditional steganography.

Use of third-party keyboards makes Android more flexible and customizable. However, we demonstrate their potential security risks by implementing a proof-of-concept keylogger that can effectively steal users' sensitive keystrokes with 81 popular websites (out of 100 tested websites). We also empirically analyzed the security behaviors of 139 keyboard applications that were available on Google Play. Our study results show that the majority of existing keyboard applications (84 out of 139) could be potentially misused as malicious keyloggers. To avoid such keylogging attacks, we discuss possible defense mechanisms.

Electronic door locks must be carefully designed to allow valid users to open (or close) a door and prevent unauthorized people from opening (or closing) the door. However, lock manufacturers have often ignored the fact that door locks can be modified by attackers in the real world. In this paper, we demonstrate that the most popular electronic door locks can easily be compromised by inserting a malicious hardware backdoor to perform unauthorized operations on the door locks. Attackers can replay a valid DC voltage pulse to open (or close) the door in an unauthorized manner or capture the user's personal identification number (PIN) used for the door lock.

Security protocols are notoriously difficult to get right, and most go through several iterations before their hidden security vulnerabilities, which are hard to detect, are triggered. To help protocol designers and developers efficiently find non-trivial bugs, we introduce SYMCONF, a practical conformance testing tool that generates high-coverage test input packets using a conformance test suite and symbolic execution. Our approach can be viewed as the combination of conformance testing and symbolic execution: 1) it first selects symbolic inputs from an existing conformance test suite; 2) it then symbolically executes a network protocol implementation with the symbolic inputs; and 3) it finally generates high-coverage test input packets using a conformance test suite. We demonstrate the feasibility of this methodology by applying SYMCONF to the generation of a stream of high quality test input packets for multiple implementations of two network protocols, the Kerberos Telnet protocol and Dynamic Host Configuration Protocol (DHCP), and discovering non-trivial security bugs in the protocols.

In modern politics, parties and individual candidates must have an online presence and usually have dedicated social media coordinators. In this context, we study the usefulness of analysing Twitter messages to identify both the characteristics of political parties and the political leaning of users. As a case study, we collected the main stream of Twitter related to the 2010 UK General Election during the associated period -- gathering around 1,150,000 messages from about 220,000 users. We examined the characteristics of the three main parties in the election and highlighted the main differences between parties. First, Lab our members were the most active and influential during the election while Conservative members were the most organized to promote their activities. Second, the websites and blogs that each political party's members supported are clearly different from those that all the other political parties' members supported. From these observations, we develop a simple and practical classification method which uses the number of Twitter messages referring to a particular political party. The experimental results showed that the proposed classification method achieved about 86% classification accuracy and outperforms other classification methods that require expensive costs for tuning classifier parameters and/or knowledge about network topology.

The problem of maximizing information diffusion through a network is a topic of considerable recent interest. A conventional problem is to select a set of any arbitrary k nodes as the initial influenced nodes so that they can effectively disseminate the information to the rest of the network. However, this model is usually unrealistic in online social networks since we cannot typically choose arbitrary nodes in the network as the initial influenced nodes. From the point of view of an individual user who wants to spread information as much as possible, a more reasonable model is to try to initially share the information with only some of its neighbours rather than a set of any arbitrary nodes; but how can these neighbours be effectively chosen? We empirically study how to design more effective neighbours selection strategies to maximize information diffusion. Our experimental results through intensive simulation on several real- world network topologies show that an effective neighbours selection strategy is to use node degree information for short-term propagation while a naive random selection is also adequate for long-term propagation to cover more than half of a network. We also discuss the effects of the number of initial activated neighbours. If we particularly select the highest degree nodes as initial activated neighbours, the number of initial activated neighbours is not an important factor at least for long-term propagation of information.

Increases in the number of TV channels requires users to spend more time to select their preferred channels since the user interaction for browsing is practically limited to the conventional remote control with a two-way scrolling button. We formally define the problem to construct the optimal channel ordering which minimizes the seek distance in selecting channels and show this problem is NP-hard. In addition, we present a reasonable heuristic to solve this problem. The proposed method constructs an efficient channel ordering by applying a hierarchical clustering algorithm based on the frequencies of switching events between channels. We demonstrate the feasibility of this method by applying a number of well-known hierarchical clustering algorithms and evaluating the number of user inputs required for selecting channels. Our experimental results show that the proposed method significantly decreases the number of user inputs compared with the conventional methods.

Mobile phones are integral to everyday life with emails, social networking, online banking and other applications; however, the wealth of private information accessible increases economic incentives for attackers. Compared with fixed networks, mobile malware can replicate through both long range messaging and short range radio technologies; the former can be filtered by the network operator but determining the best method of containing short range malware is an open problem. While global software updates are sometimes possible, they are often not practical. An alternative and more efficient strategy is to distribute the patch to the key nodes so that they can opportunistically disseminate it to the rest of the network via short range encounters; but how can these key nodes be identified in a highly dynamic network topology? In this paper, we address these questions by presenting Socio- Temporal Opportunistic Patching (STOP), a two-tier predictive mobile malware containment system: devices collect co-location data in a decentralized manner and report to a central server which processes and targets delivery of hot fixes to a small subset of k devices at runtime; in turn mobile devices spread the patch opportunistically. The STOP system is underpinned by a recent theoretical framework for analysing dynamic networks that takes into account temporal information of links. Using empirical contact traces, we find firstly, the top-k ranking temporal centrality nodes are highly correlated with past time windows; and secondly, simple prediction functions can be designed to select the set of top-k nodes that are optimal for patch spreading.

A mobile ad hoc network (MANET) is a decentralized network of mobile nodes. Due to the broadcast nature of radio transmissions, communication in MANETs is more susceptible to malicious traffic analysis. An interesting problem is how to thwart malicious traffic analysis. Most anonymous communication protocols are based on the pseudonyms of mobile nodes. However, conventional pseudonym schemes have some limitations such as collisions of pseudonyms and high computational complexity due to the use of cryptographic hash functions. Collisions of identities are not desirable since they are the main causes for reduced effective bandwidth, increased energy consumption and non-deterministic data delivery. In this paper, we propose a new collision-free pseudonym scheme to enable anonymous communication. In our approach, each node generates pseudonyms by using a permutation matrix without collisions. The challenging issue is how to store the overall permutation matrix. It is practically hard to assume that mobile nodes maintain the permutation matrix due to the limitation of resources. Therefore we design the online computation of each node's own pseudonym without loading the overall matrix.