arXiv subjects

Tingmin Wu

Publications and source records attributed to Tingmin Wu.

17 recordsLinked to original sources

arXiv2026

GRIDEX: Grid-Grounded Forensic Explanations for Deepfake Spectrogram Analysis

The advancement of speech generation technologies has made artificial speech increasingly realistic. Although modern classification models can achieve high accuracy when it comes to deepfake detection, they do not produce evidences such as indicating where spoof cues appear in the spectrogram and what they imply acoustically, limiting their usefulness in forensic settings. Manual analysis of full spectrograms is resource-intensive, so evidence should narrow attention to the most diagnostic regions. Moreover, existing explainability methods have limited capabilities in connecting contextual attributes to localized evidence, making explanations harder to verify. To overcome this limitation, we propose GRIDEX, a pipeline that, when given a deepfake spectrogram, generates forensic explanations of its anomalies. The pipeline (i) selects top-K anomalous regions in the spectrogram and (ii) produces an explanation for each anomaly. The explanations follow a schema of categorical acoustic fields, including temporal, spectral, phonetic information and interpretation text. To our knowledge, this is the first framework to generate structured forensic explanations using regional grounding for deepfake spectrograms. GRIDEX is trained with a two-stage learning paradigm that combines supervised fine-tuning (SFT) with Group Relative Policy Optimization (GRPO). Experiments on our dataset show improved artifact localization and explanation quality over strong vision-language model (VLM) baselines. The dataset and code will be released upon publication.

cs.SD↗

arXiv2026

Agents at Risk: How Users Unwittingly Undermine LLM Safety

Large language model (LLM)-based agents are increasingly deployed in applications, such as trip-planning agents and web-use agents, to perform complex planning and execution tasks. Prior work has shown that LLM-based agents are vulnerable to context confusion, where external adversarial content incorporated into the agent's reasoning context may be treated as task-relevant constraints. However, external malicious content can enter the agent context via channels beyond retrieval. In this work, we introduce the User-Relayed Context Manipulation (UReCoM) attack, in which attackers manipulate benign users into relaying adversarial content within user requests, thereby relocating external adversarial content into user-provided task context. Our experimental evaluation shows that UReCoM outperforms five prompt-injection baselines (naive, context ignoring, fake completion, escape-character attacks, and combined attacks) under prevention-based (Sandwich, StruQ, and SecAlign) and detection-based defenses (Perplexity detection, DataSentinel, and CausalArmor). Additionally, UReCoM shows that LLMs can reject explicit malicious instructions more reliably than they can identify adversarial task entities, such as promotion codes, embedded within user requests. On 12 commercial LLM-based agents, we find that validation of adversarial task entities is largely prompt-driven rather than default, highlighting a design flaw in current agent frameworks. These results indicate that current defenses and deployed agents remain insufficient against user-relayed context manipulations, highlighting the need for task-entity-level prevention and default safety verification in LLM-based agents.

cs.CR↗

arXiv2025

Can Current Detectors Catch Face-to-Voice Deepfake Attacks?

The rapid advancement of generative models has enabled the creation of increasingly stealthy synthetic voices, commonly referred to as audio deepfakes. A recent technique, FOICE [USENIX'24], demonstrates a particularly alarming capability: generating a victim's voice from a single facial image, without requiring any voice sample. By exploiting correlations between facial and vocal features, FOICE produces synthetic voices realistic enough to bypass industry-standard authentication systems, including WeChat Voiceprint and Microsoft Azure. This raises serious security concerns, as facial images are far easier for adversaries to obtain than voice samples, dramatically lowering the barrier to large-scale attacks. In this work, we investigate two core research questions: (RQ1) can state-of-the-art audio deepfake detectors reliably detect FOICE-generated speech under clean and noisy conditions, and (RQ2) whether fine-tuning these detectors on FOICE data improves detection without overfitting, thereby preserving robustness to unseen voice generators such as SpeechT5. Our study makes three contributions. First, we present the first systematic evaluation of FOICE detection, showing that leading detectors consistently fail under both standard and noisy conditions. Second, we introduce targeted fine-tuning strategies that capture FOICE-specific artifacts, yielding significant accuracy improvements. Third, we assess generalization after fine-tuning, revealing trade-offs between specialization to FOICE and robustness to unseen synthesis pipelines. These findings expose fundamental weaknesses in today's defenses and motivate new architectures and training protocols for next-generation audio deepfake detection.

cs.CR↗

arXiv2025

MulVuln: Enhancing Pre-trained LMs with Shared and Language-Specific Knowledge for Multilingual Vulnerability Detection

Software vulnerabilities (SVs) pose a critical threat to safety-critical systems, driving the adoption of AI-based approaches such as machine learning and deep learning for software vulnerability detection. Despite promising results, most existing methods are limited to a single programming language. This is problematic given the multilingual nature of modern software, which is often complex and written in multiple languages. Current approaches often face challenges in capturing both shared and language-specific knowledge of source code, which can limit their performance on diverse programming languages and real-world codebases. To address this gap, we propose MULVULN, a novel multilingual vulnerability detection approach that learns from source code across multiple languages. MULVULN captures both the shared knowledge that generalizes across languages and the language-specific knowledge that reflects unique coding conventions. By integrating these aspects, it achieves more robust and effective detection of vulnerabilities in real-world multilingual software systems. The rigorous and extensive experiments on the real-world and diverse REEF dataset, consisting of 4,466 CVEs with 30,987 patches across seven programming languages, demonstrate the superiority of MULVULN over thirteen effective and state-of-the-art baselines. Notably, MULVULN achieves substantially higher F1-score, with improvements ranging from 1.45% to 23.59% compared to the baseline methods.

cs.CR↗

arXiv2025

SoK: Systematizing Generation, Characteristics, and Defenses in LLM-Generated Phishing

The rapid advancement of Large Language Models (LLMs), with their growing abuse in phishing, has enabled phishing content, including deceptive pretexts and other persuasive elements, to be generated at a scale difficult to achieve manually. This growing misuse of LLMs in phishing raises questions about potential LLM-driven changes in phishing characteristics, associated security implications for users, and the resulting challenges to existing defenses. While prior research has examined individual dimensions, including threats of LLM abuse, LLM-driven phishing, user susceptibility to phishing, and phishing defenses, these efforts remain fragmented and have not been consolidated into a comprehensive understanding of how these dimensions interrelate. To address this research gap, we provide a systematic examination of the LLM-generated phishing landscape, including a taxonomy of LLM manipulation methods, characteristics associated with LLM-generated phishing threats, and a taxonomy of defenses aligned with these manipulation methods. We also benchmark five academic and five industrial phishing detectors across datasets associated with different LLM-based generation methods. We further extract insights and research gaps that suggest promising directions for future research in this growing area. Our findings underscore the need for countermeasures that are evaluated across LLM generators and manipulation strategies. Our work provides a systematic foundation for studying LLM-generated phishing, enabling more consistent comparison and evaluation across the community.

cs.CR↗

arXiv2025

LLMs Are Not Yet Ready for Deepfake Image Detection

The growing sophistication of deepfakes presents substantial challenges to the integrity of media and the preservation of public trust. Concurrently, vision-language models (VLMs), large language models enhanced with visual reasoning capabilities, have emerged as promising tools across various domains, sparking interest in their applicability to deepfake detection. This study conducts a structured zero-shot evaluation of four prominent VLMs: ChatGPT, Claude, Gemini, and Grok, focusing on three primary deepfake types: faceswap, reenactment, and synthetic generation. Leveraging a meticulously assembled benchmark comprising authentic and manipulated images from diverse sources, we evaluate each model's classification accuracy and reasoning depth. Our analysis indicates that while VLMs can produce coherent explanations and detect surface-level anomalies, they are not yet dependable as standalone detection systems. We highlight critical failure modes, such as an overemphasis on stylistic elements and vulnerability to misleading visual patterns like vintage aesthetics. Nevertheless, VLMs exhibit strengths in interpretability and contextual analysis, suggesting their potential to augment human expertise in forensic workflows. These insights imply that although general-purpose models currently lack the reliability needed for autonomous deepfake detection, they hold promise as integral components in hybrid or human-in-the-loop detection frameworks.

cs.CV↗

arXiv2025

ThreatModeling-LLM: Automating Threat Modeling using Large Language Models for Banking System

Threat modeling is a crucial component of cybersecurity, particularly for industries such as banking, where the security of financial data is paramount. Traditional threat modeling approaches require expert intervention and manual effort, often leading to inefficiencies and human error. The advent of Large Language Models (LLMs) offers a promising avenue for automating these processes, enhancing both efficiency and efficacy. However, this transition is not straightforward due to three main challenges: (1) the lack of publicly available, domain-specific datasets, (2) the need for tailored models to handle complex banking system architectures, and (3) the requirement for real-time, adaptive mitigation strategies that align with compliance standards like NIST 800-53. In this paper, we introduce ThreatModeling-LLM, a novel and adaptable framework that automates threat modeling for banking systems using LLMs. ThreatModeling-LLM operates in three stages: 1) dataset creation, 2) prompt engineering and 3) model fine-tuning. We first generate a benchmark dataset using Microsoft Threat Modeling Tool (TMT). Then, we apply Chain of Thought (CoT) and Optimization by PROmpting (OPRO) on the pre-trained LLMs to optimize the initial prompt. Lastly, we fine-tune the LLM using Low-Rank Adaptation (LoRA) based on the benchmark dataset and the optimized prompt to improve the threat identification and mitigation generation capabilities of pre-trained LLMs.

cs.CR↗

arXiv2025

PEEK: Phishing Evolution Framework for Phishing Generation and Evolving Pattern Analysis using Large Language Models

Phishing remains a pervasive cyber threat, as attackers craft deceptive emails to lure victims into revealing sensitive information. While Artificial Intelligence (AI), in particular, deep learning, has become a key component in defending against phishing attacks, these approaches face critical limitations. The scarcity of publicly available, diverse, and updated data, largely due to privacy concerns, constrains detection effectiveness. As phishing tactics evolve rapidly, models trained on limited, outdated data struggle to detect new, sophisticated deception strategies, leaving systems and people vulnerable to an ever-growing array of attacks. We propose the first Phishing Evolution FramEworK (PEEK) for augmenting phishing email datasets with respect to quality and diversity, and analyzing changing phishing patterns for detection to adapt to updated phishing attacks. Specifically, we integrate large language models (LLMs) into the process of adversarial training to enhance the performance of the generated dataset and leverage persuasion principles in a recurrent framework to facilitate the understanding of changing phishing strategies. PEEK raises the proportion of usable phishing samples from 21.4% to 84.8%, surpassing existing works that rely on prompting and fine-tuning LLMs. The phishing datasets provided by PEEK, with evolving phishing patterns, outperform the other two available LLM-generated phishing email datasets in improving detection robustness. PEEK phishing boosts detectors' accuracy to over 88% and reduces adversarial sensitivity by up to 70%, still maintaining 70% detection accuracy against adversarial attacks.

cs.CR↗

arXiv2024

SAFE: Advancing Large Language Models in Leveraging Semantic and Syntactic Relationships for Software Vulnerability Detection

Software vulnerabilities (SVs) have emerged as a prevalent and critical concern for safety-critical security systems. This has spurred significant advancements in utilizing AI-based methods, including machine learning and deep learning, for software vulnerability detection (SVD). While AI-based methods have shown promising performance in SVD, their effectiveness on real-world, complex, and diverse source code datasets remains limited in practice. To tackle this challenge, in this paper, we propose a novel framework that enhances the capability of large language models to learn and utilize semantic and syntactic relationships from source code data for SVD. As a result, our approach can enable the acquisition of fundamental knowledge from source code data while adeptly utilizing crucial relationships, i.e., semantic and syntactic associations, to effectively address the software vulnerability detection (SVD) problem. The rigorous and extensive experimental results on three real-world challenging datasets (i.e., ReVeal, D2A, and Devign) demonstrate the superiority of our approach over the effective and state-of-the-art baselines. In summary, on average, our SAFE approach achieves higher performances from 4.79% to 9.15% for F1-measure and from 16.93% to 21.70% for Recall compared to the baselines across all datasets used.

cs.SE↗

arXiv2024

An Innovative Information Theory-based Approach to Tackle and Enhance The Transparency in Phishing Detection

Phishing attacks have become a serious and challenging issue for detection, explanation, and defense. Despite more than a decade of research on phishing, encompassing both technical and non-technical remedies, phishing continues to be a serious problem. Nowadays, AI-based phishing detection stands out as one of the most effective solutions for defending against phishing attacks by providing vulnerability (i.e., phishing or benign) predictions for the data. However, it lacks explainability in terms of providing comprehensive interpretations for the predictions, such as identifying the specific information that causes the data to be classified as phishing. To this end, we propose an innovative deep learning-based approach for email (the most common phishing way) phishing attack localization. Our method can not only predict the vulnerability of the email data but also automatically learn and figure out the most important and phishing-relevant information (i.e., sentences) in the phishing email data where the selected information indicates useful and concise explanations for the vulnerability. The rigorous experiments on seven real-world diverse email datasets show the effectiveness and advancement of our proposed method in selecting crucial information, offering concise explanations (by successfully figuring out the most important and phishing-relevant information) for the vulnerability of the phishing email data. Particularly, our method achieves a significantly higher performance, ranging from approximately 1.5% to 3.5%, compared to state-of-the-art baselines, as measured by the combined average performance of two main metrics Label-Accuracy and Cognitive-True-Positive.

cs.CR↗

arXiv2024

Deep Learning-Based Out-of-distribution Source Code Data Identification: How Far Have We Gone?

Software vulnerabilities (SVs) have become a common, serious, and crucial concern to safety-critical security systems. That leads to significant progress in the use of AI-based methods for software vulnerability detection (SVD). In practice, although AI-based methods have been achieving promising performances in SVD and other domain applications (e.g., computer vision), they are well-known to fail in detecting the ground-truth label of input data (referred to as out-of-distribution, OOD, data) lying far away from the training data distribution (i.e., in-distribution, ID). This drawback leads to serious issues where the models fail to indicate when they are likely mistaken. To address this problem, OOD detectors (i.e., determining whether an input is ID or OOD) have been applied before feeding the input data to the downstream AI-based modules. While OOD detection has been widely designed for computer vision and medical diagnosis applications, automated AI-based techniques for OOD source code data detection have not yet been well-studied and explored. To this end, in this paper, we propose an innovative deep learning-based approach addressing the OOD source code data identification problem. Our method is derived from an information-theoretic perspective with the use of innovative cluster-contrastive learning to effectively learn and leverage source code characteristics, enhancing data representation learning for solving the problem. The rigorous and comprehensive experiments on real-world source code datasets show the effectiveness and advancement of our approach compared to state-of-the-art baselines by a wide margin. In short, on average, our method achieves a significantly higher performance from around 15.27%, 7.39%, and 4.93% on the FPR, AUROC, and AUPR measures, respectively, in comparison with the baselines.

cs.CR↗

arXiv2023

Copyright Protection and Accountability of Generative AI:Attack, Watermarking and Attribution

Generative AI (e.g., Generative Adversarial Networks - GANs) has become increasingly popular in recent years. However, Generative AI introduces significant concerns regarding the protection of Intellectual Property Rights (IPR) (resp. model accountability) pertaining to images (resp. toxic images) and models (resp. poisoned models) generated. In this paper, we propose an evaluation framework to provide a comprehensive overview of the current state of the copyright protection measures for GANs, evaluate their performance across a diverse range of GAN architectures, and identify the factors that affect their performance and future research directions. Our findings indicate that the current IPR protection methods for input images, model watermarking, and attribution networks are largely satisfactory for a wide range of GANs. We highlight that further attention must be directed towards protecting training sets, as the current approaches fail to provide robust IPR protection and provenance tracing on training sets.

cs.LG↗

arXiv2023

RAIDER: Reinforcement-aided Spear Phishing Detector

Spear Phishing is a harmful cyber-attack facing business and individuals worldwide. Considerable research has been conducted recently into the use of Machine Learning (ML) techniques to detect spear-phishing emails. ML-based solutions may suffer from zero-day attacks; unseen attacks unaccounted for in the training data. As new attacks emerge, classifiers trained on older data are unable to detect these new varieties of attacks resulting in increasingly inaccurate predictions. Spear Phishing detection also faces scalability challenges due to the growth of the required features which is proportional to the number of the senders within a receiver mailbox. This differs from traditional phishing attacks which typically perform only a binary classification between phishing and benign emails. Therefore, we devise a possible solution to these problems, named RAIDER: Reinforcement AIded Spear Phishing DEtectoR. A reinforcement-learning based feature evaluation system that can automatically find the optimum features for detecting different types of attacks. By leveraging a reward and penalty system, RAIDER allows for autonomous features selection. RAIDER also keeps the number of features to a minimum by selecting only the significant features to represent phishing emails and detect spear-phishing attacks. After extensive evaluation of RAIDER over 11,000 emails and across 3 attack scenarios, our results suggest that using reinforcement learning to automatically identify the significant features could reduce the dimensions of the required features by 55% in comparison to existing ML-based systems. It also improves the accuracy of detecting spoofing attacks by 4% from 90% to 94%. In addition, RAIDER demonstrates reasonable detection accuracy even against a sophisticated attack named Known Sender in which spear-phishing emails greatly resemble those of the impersonated sender.

cs.CR↗

arXiv2022

Profiler: Profile-Based Model to Detect Phishing Emails

Email phishing has become more prevalent and grows more sophisticated over time. To combat this rise, many machine learning (ML) algorithms for detecting phishing emails have been developed. However, due to the limited email data sets on which these algorithms train, they are not adept at recognising varied attacks and, thus, suffer from concept drift; attackers can introduce small changes in the statistical characteristics of their emails or websites to successfully bypass detection. Over time, a gap develops between the reported accuracy from literature and the algorithm's actual effectiveness in the real world. This realises itself in frequent false positive and false negative classifications. To this end, we propose a multidimensional risk assessment of emails to reduce the feasibility of an attacker adapting their email and avoiding detection. This horizontal approach to email phishing detection profiles an incoming email on its main features. We develop a risk assessment framework that includes three models which analyse an email's (1) threat level, (2) cognitive manipulation, and (3) email type, which we combine to return the final risk assessment score. The Profiler does not require large data sets to train on to be effective and its analysis of varied email features reduces the impact of concept drift. Our Profiler can be used in conjunction with ML approaches, to reduce their misclassifications or as a labeller for large email data sets in the training stage. We evaluate the efficacy of the Profiler against a machine learning ensemble using state-of-the-art ML algorithms on a data set of 9000 legitimate and 900 phishing emails from a large Australian research organisation. Our results indicate that the Profiler's mitigates the impact of concept drift, and delivers 30% less false positive and 25% less false negative email classifications over the ML ensemble's approach.

cs.CR↗

arXiv2022

Email Summarization to Assist Users in Phishing Identification

Cyber-phishing attacks recently became more precise, targeted, and tailored by training data to activate only in the presence of specific information or cues. They are adaptable to a much greater extent than traditional phishing detection. Hence, automated detection systems cannot always be 100% accurate, increasing the uncertainty around expected behavior when faced with a potential phishing email. On the other hand, human-centric defence approaches focus extensively on user training but face the difficulty of keeping users up to date with continuously emerging patterns. Therefore, advances in analyzing the content of an email in novel ways along with summarizing the most pertinent content to the recipients of emails is a prospective gateway to furthering how to combat these threats. Addressing this gap, this work leverages transformer-based machine learning to (i) analyze prospective psychological triggers, to (ii) detect possible malicious intent, and (iii) create representative summaries of emails. We then amalgamate this information and present it to the user to allow them to (i) easily decide whether the email is "phishy" and (ii) self-learn advanced malicious patterns.

cs.CR↗

arXiv2020

Catering to Your Concerns: Automatic Generation of Personalised Security-Centric Descriptions for Android Apps

Android users are increasingly concerned with the privacy of their data and security of their devices. To improve the security awareness of users, recent automatic techniques produce security-centric descriptions by performing program analysis. However, the generated text does not always address users' concerns as they are generally too technical to be understood by ordinary users. Moreover, different users have varied linguistic preferences, which do not match the text. Motivated by this challenge, we develop an innovative scheme to help users avoid malware and privacy-breaching apps by generating security descriptions that explain the privacy and security related aspects of an Android app in clear and understandable terms. We implement a prototype system, PERSCRIPTION, to generate personalised security-centric descriptions that automatically learn users' security concerns and linguistic preferences to produce user-oriented descriptions. We evaluate our scheme through experiments and user studies. The results clearly demonstrate the improvement on readability and users' security awareness of PERSCRIPTION's descriptions compared to existing description generators.

cs.CR↗

arXiv2020

Analysis of Trending Topics and Text-based Channels of Information Delivery in Cybersecurity

Computer users are generally faced with difficulties in making correct security decisions. While an increasingly fewer number of people are trying or willing to take formal security training, online sources including news, security blogs, and websites are continuously making security knowledge more accessible. Analysis of cybersecurity texts can provide insights into the trending topics and identify current security issues as well as how cyber attacks evolve over time. These in turn can support researchers and practitioners in predicting and preparing for these attacks. Comparing different sources may facilitate the learning process for normal users by persisting the security knowledge gained from different cybersecurity context. Prior studies neither systematically analysed the wide-range of digital sources nor provided any standardisation in analysing the trending topics from recent security texts. Although LDA has been widely adopted in topic generation, its generated topics cannot cover the cybersecurity concepts completely and considerably overlap. To address this issue, we propose a semi-automated classification method to generate comprehensive security categories instead of LDA-generated topics. We further compare the identified 16 security categories across different sources based on their popularity and impact. We have revealed several surprising findings. (1) The impact reflected from cyber-security texts strongly correlates with the monetary loss caused by cybercrimes. (2) For most categories, security blogs share the largest popularity and largest absolute/relative impact over time. (3) Websites deliver security information without caring about timeliness much, where one third of the articles do not specify the date and the rest have a time lag in posting emerging security issues.

cs.CR↗