arXiv subjects

Murtuza Jadliwala

Publications and source records attributed to Murtuza Jadliwala.

At least 19 recordsLinked to original sources

arXiv2026

Reducing Hallucinated Transcripts in Whisper via Hallucination Space Projection

Whisper is a widely used foundation model for automatic speech recognition (ASR), but its generative decoder can produce fluent hallucinated transcripts for inputs containing little or no speech. We propose a training-free, inference-time method to reduce these hallucinations using low-rank projection of decoder activations. A compact hallucination-associated subspace is estimated from non-speech calibration data, and decoder hidden states are projected away from this subspace during inference. We evaluate two variants: always-on, which applies projection to all inputs, and gated, which applies it only when Whisper predicts that an input is likely non-speech. Across non-speech benchmarks, always-on projection reduces average hallucination rate (HR) from 31.31% to 2.44%, a 92.21% relative reduction, while gated projection reduces HR to 3.74%, an 88.05% relative reduction, with lower false rejection of genuine speech. On LibriSpeech, gated projection increases absolute word error rate (WER) by 0.33-4.39 percentage points and yields false-rejection rates (FRR) of 0.41--9.97% across model and split settings. These results show that low-rank activation projection can substantially suppress Whisper hallucinations without retraining, while providing a controllable trade-off between hallucination suppression and speech recognition performance.

cs.AI↗

arXiv2026

Beyond Single-Vulnerability Evaluation: Closing the Engineering Decision Gap Between C Retrofits and Native Safety

While decades of research have produced numerous retrofitted memory-safety protections for C, these mechanisms are almost exclusively evaluated in isolation, targeting specific vulnerability classes. This siloed evaluation paradigm leaves practitioners without a clear understanding of the cumulative performance costs, interoperability conflicts, and protection gaps that arise when layering defenses to achieve comprehensive safety. This paper presents a new evaluation paradigm that benchmarks natively memory-safe languages like Rust and Go against compounded C retrofits. Using standardized cross-language tasks, we evaluate the performance and protection tradeoffs of state-of-the-art mechanisms when deployed in combination. Our results demonstrate that layered C defenses incur compounding and workload-dependent performance penalties, can suffer from fundamental architectural incompatibilities, and fall short of the protection scope provided by native memory-safe languages. These findings expose a critical engineering decision gap where the true cost of backporting safety to C remains hidden from practitioners. We argue for a fundamental shift in memory-safety research: moving away from isolated evaluation toward holistic, comparative frameworks that inform the high-stakes choice between retrofitting legacy codebases and migrating to modern, safe languages.

cs.PL↗

arXiv2026

Trust but Verify? Uncovering the Security Debt of Autonomous Coding Agents

The increasing adoption of autonomous coding agents accelerates software development but also introduces scoped security risks within high-impact file paths that can outpace traditional human review capacity. While prior research has primarily evaluated these systems in terms of functional correctness and productivity, this paper presents a large-scale empirical study using the AIDev dataset to systematically characterize security code smells in agent-generated pull requests (PRs). Through a combination of a validated LLM-as-a-judge framework and manual qualitative analysis, we identify and classify security misconfigurations across 16,112 file changes spanning 4,022 pull requests. Our results reveal that 38.9% of agent-generated PRs contain at least one security smell, with supply chain integrity issues accounting for 82.3% of all detected security smells. Furthermore, hard-coded credentials constitute 99.6% of all critical-severity security smells. Crucially, we find that human collaborators are responsible for introducing 67.6% of genuine leaked secrets within these agent-assisted workflows, while existing automated and human review processes fail to detect 81.1% of these credentials prior to integration. These findings highlight substantial security risks in agent-assisted software development workflows and suggest a potential reduction in developer vigilance. They also underscore the urgent need for context-aware security guardrails implemented directly at the point of human-AI collaboration.

cs.CR↗

arXiv2026

ThermalTap: Passive Application Fingerprinting in VR Headsets via Thermal Side Channels

Standalone virtual reality (VR) headsets process highly sensitive personal, professional, and health-related data, yet their susceptibility to non-contact physical side channels remains largely unexplored. Existing side-channel attacks typically require malicious software execution or physical access to peripherals, making them conspicuous and potentially patchable. This paper introduces ThermalTap, the first passive, non-contact side-channel attack that fingerprints VR applications solely from the long-wave infrared (LWIR) radiation emitted by the headset chassis. By treating a headset's thermal signature as a high-fidelity proxy for internal computational workloads, ThermalTap enables remote application inference at meter-scale distances without any device interaction. To achieve robust performance in real-world settings, the system combines a commodity thermal camera with a multi-modal sensor suite (capturing ambient temperature, humidity, and airflow) to normalize environmental noise. We evaluate ThermalTap using six applications across three commercial standalone headsets. In indoor settings, ThermalTap identifies applications with over 90% accuracy using only 10 seconds of thermal camera data. Under outdoor conditions, with longer session-level observations, several applications remain identifiable despite environmental variability, with the strongest outdoor application reaching 81% accuracy. Our findings establish thermal radiation as a fundamental and unavoidable privacy risk for immersive systems, exposing a critical security gap that bypasses current software-level protections and physical access controls.

cs.CR↗

arXiv2026

LLM Ghostbusters: Surgical Hallucination Suppression via Adaptive Unlearning

Hallucinations, outputs that sound plausible but are factually incorrect, remain an open challenge for deployed LLMs. In code generation, models frequently hallucinate non-existent software packages, recommending imports and installation commands for fictional libraries. This creates a critical supply-chain vulnerability: an attacker can proactively register such packages on public registries with malicious payloads that are subsequently installed and executed by developers or autonomous agents, a class of package confusion attack known as slopsquatting. Once a model is deployed, mitigating this failure mode is difficult: full retraining is costly, and existing approaches either cause severe degradation of model utility or rely on a pre-specified forget-set, an assumption that does not apply to the unbounded space of hallucinations. To address this problem, we present Adaptive Unlearning (AU), a post-deployment framework that surgically suppresses hallucinations while preserving general model utility. AU introduces a hybrid token-level objective that simultaneously reinforces valid outputs and suppresses hallucinated ones. Combined with an adaptive discovery loop that continuously surfaces new hallucination-inducing contexts without human supervision, AU enables generalization to unseen prompts and hallucinations. We demonstrate that AU reduces package hallucination rates by 81%, corresponding to a substantial reduction in slopsquatting attack surface, while maintaining performance on standard coding benchmarks. Our analysis shows that distributional changes are concentrated on package-related generations, leaving general coding behavior largely unaffected and confirming that AU's effect is isolated to the targeted distribution. AU operates entirely on model-generated data, requires no human annotation, and generalizes across domains.

cs.CR↗

arXiv2026

Off-The-Shelf Image-to-Image Models Are All You Need To Defeat Image Protection Schemes

Advances in Generative AI (GenAI) have led to the development of various protection strategies to prevent the unauthorized use of images. These methods rely on adding imperceptible protective perturbations to images to thwart misuse such as style mimicry or deepfake manipulations. Although previous attacks on these protections required specialized, purpose-built methods, we demonstrate that this is no longer necessary. We show that off-the-shelf image-to-image GenAI models can be repurposed as generic ``denoisers" using a simple text prompt, effectively removing a wide range of protective perturbations. Across 8 case studies spanning 6 diverse protection schemes, our general-purpose attack not only circumvents these defenses but also outperforms existing specialized attacks while preserving the image's utility for the adversary. Our findings reveal a critical and widespread vulnerability in the current landscape of image protection, indicating that many schemes provide a false sense of security. We stress the urgent need to develop robust defenses and establish that any future protection mechanism must be benchmarked against attacks from off-the-shelf GenAI models. Code is available in this repository: https://github.com/mlsecviswanath/img2imgdenoiser

cs.CV↗

arXiv2026

Prompt and Circumstances: Evaluating the Efficacy of Human Prompt Inference in AI-Generated Art

The emerging field of AI-generated art has witnessed the rise of prompt marketplaces, where creators can purchase, sell, or share prompts to generate unique artworks. These marketplaces often assert ownership over prompts, claiming them as intellectual property. This paper investigates whether concealed prompts sold on prompt marketplaces can be considered bona fide intellectual property, given that humans and AI tools may be able to infer the prompts based on publicly advertised sample images accompanying each prompt on sale. Specifically, our study aims to assess (i) how accurately humans can infer the original prompt solely by examining an AI-generated image, with the goal of generating images similar to the original image, and (ii) the possibility of improving upon individual human and AI prompt inferences by crafting combined human and AI prompts with the help of a large language model. Although previous research has explored AI-driven prompt inference and protection strategies, our work is the first to incorporate a human subject study and examine collaborative human-AI prompt inference in depth. Our findings indicate that while prompts inferred by humans and prompts inferred through a combined human and AI effort can generate images with a moderate level of similarity, they are not as successful as using the original prompt. Moreover, combining human- and AI-inferred prompts using our suggested merging techniques did not improve performance over purely human-inferred prompts.

cs.CR↗

arXiv2026

VocalBridge: Latent Diffusion-Bridge Purification for Defeating Perturbation-Based Voiceprint Defenses

The rapid advancement of speech synthesis technologies, including text-to-speech (TTS) and voice conversion (VC), has intensified security and privacy concerns related to voice cloning. Recent defenses attempt to prevent unauthorized cloning by embedding protective perturbations into speech to obscure speaker identity while maintaining intelligibility. However, adversaries can apply advanced purification techniques to remove these perturbations, recover authentic acoustic characteristics, and regenerate cloneable voices. Despite the growing realism of such attacks, the robustness of existing defenses under adaptive purification remains insufficiently studied. Most existing purification methods are designed to counter adversarial noise in automatic speech recognition (ASR) systems rather than speaker verification or voice cloning pipelines. As a result, they fail to suppress the fine-grained acoustic cues that define speaker identity and are often ineffective against speaker verification attacks (SVA). To address these limitations, we propose Diffusion-Bridge (VocalBridge), a purification framework that learns a latent mapping from perturbed to clean speech in the EnCodec latent space. Using a time-conditioned 1D U-Net with a cosine noise schedule, the model enables efficient, transcript-free purification while preserving speaker-discriminative structure. We further introduce a Whisper-guided phoneme variant that incorporates lightweight temporal guidance without requiring ground-truth transcripts. Experimental results show that our approach consistently outperforms existing purification methods in recovering cloneable voices from protected speech. Our findings demonstrate the fragility of current perturbation-based defenses and highlight the need for more robust protection mechanisms against evolving voice-cloning and speaker verification threats.

cs.SD↗

arXiv2025

Linguistic Hooks: Investigating The Role of Language Triggers in Phishing Emails Targeting African Refugees and Students

Phishing and sophisticated email-based social engineering attacks disproportionately affect vulnerable populations, such as refugees and immigrant students. However, these groups remain understudied in cybersecurity research. This gap in understanding, coupled with their exclusion from broader security and privacy policies, increases their susceptibility to phishing and widens the digital security divide between marginalized and non-marginalized populations. To address this gap, we first conducted digital literacy workshops with newly resettled African refugee populations (n = 48) in the US to improve their understanding of how to safeguard sensitive and private information. Following the workshops, we conducted a real-world phishing deception study using carefully designed emails with linguistic cues for three participant groups: a subset of the African US-refugees recruited from the digital literacy workshops (n = 19), African immigrant students in the US (n = 142), and a control group of monolingual US-born students (n = 184). Our findings indicate that while digital literacy training for refugees improves awareness of safe cybersecurity practices, recently resettled African US-refugees still face significant challenges due to low digital literacy skills and limited English proficiency. This often leads them to ignore or fail to recognize phishing emails as phishing. Both African immigrant students and US-born students showed greater caution, though instances of data disclosure remained prevalent across groups. Our findings highlight, irrespective of literacy, the need to be trained to think critically about digital security. We conclude by discussing how the security and privacy community can better include marginalized populations in policy making and offer recommendations for designing equitable, inclusive cybersecurity initiatives.

cs.CY↗

arXiv2025

To WASM or Not to WASM: Evaluation of Browser Fingerprinting Defenses Under WASM based Obfuscation

Browser fingerprinting defenses have historically focused on detecting JavaScript(JS)-based tracking techniques. However, the widespread adoption of WebAssembly (WASM) introduces a potential blind spot, as adversaries can convert JS to WASM's low-level binary format to obfuscate malicious logic. This paper presents the first systematic evaluation of how such WASM-based obfuscation impacts the robustness of modern fingerprinting defenses. We develop an automated pipeline that translates real-world JS fingerprinting scripts into functional WASM-obfuscated variants and test them against two classes of defenses: state-of-the-art detectors in research literature and commercial, in-browser tools. Our findings reveal a notable divergence: detectors proposed in the research literature that rely on feature-based analysis of source code show moderate vulnerability, stemming from outdated datasets or a lack of WASM compatibility. In contrast, defenses such as browser extensions and native browser features remained completely effective, as their API-level interception is agnostic to the script's underlying implementation. These results highlight a gap between academic and practical defense strategies and offer insights into strengthening detection approaches against WASM-based obfuscation, while also revealing opportunities for more evasive techniques in future attacks.

cs.CR↗

arXiv2025

Optimus: A Robust Defense Framework for Mitigating Toxicity while Fine-Tuning Conversational AI

Customizing Large Language Models (LLMs) on untrusted datasets poses severe risks of injecting toxic behaviors. In this work, we introduce Optimus, a novel defense framework designed to mitigate fine-tuning harms while preserving conversational utility. Unlike existing defenses that rely heavily on precise toxicity detection or restrictive filtering, Optimus addresses the critical challenge of ensuring robust mitigation even when toxicity classifiers are imperfect or biased. Optimus integrates a training-free toxicity classification scheme that repurposes the safety alignment of commodity LLMs, and employs a dual-strategy alignment process combining synthetic "healing data" with Direct Preference Optimization (DPO) to efficiently steer models toward safety. Extensive evaluations demonstrate that Optimus mitigates toxicity even when relying on extremely biased classifiers (with up to 85% degradation in Recall). Optimus outperforms the state-of-the-art defense StarDSS and exhibits strong resilience against adaptive adversarial and jailbreak attacks. Our source code and datasets are available at https://github.com/secml-lab-vt/Optimus

cs.CR↗

arXiv2025

A Picture is Worth a Thousand Prompts? Efficacy of Iterative Human-Driven Prompt Refinement in Image Regeneration Tasks

With AI-generated content becoming ubiquitous across the web, social media, and other digital platforms, it is vital to examine how such content are inspired and generated. The creation of AI-generated images often involves refining the input prompt iteratively to achieve desired visual outcomes. This study focuses on the relatively underexplored concept of image regeneration using AI, in which a human operator attempts to closely recreate a specific target image by iteratively refining their prompt. Image regeneration is distinct from normal image generation, which lacks any predefined visual reference. A separate challenge lies in determining whether existing image similarity metrics (ISMs) can provide reliable, objective feedback in iterative workflows, given that we do not fully understand if subjective human judgments of similarity align with these metrics. Consequently, we must first validate their alignment with human perception before assessing their potential as a feedback mechanism in the iterative prompt refinement process. To address these research gaps, we present a structured user study evaluating how iterative prompt refinement affects the similarity of regenerated images relative to their targets, while also examining whether ISMs capture the same improvements perceived by human observers. Our findings suggest that incremental prompt adjustments substantially improve alignment, verified through both subjective evaluations and quantitative measures, underscoring the broader potential of iterative workflows to enhance generative AI content creation across various application domains.

cs.AI↗

arXiv2025

ScooterLab: A Programmable and Participatory Sensing Research Testbed using Micromobility Vehicles

Micromobility vehicles, such as e-scooters, are increasingly popular in urban communities but present significant challenges in terms of road safety, user privacy, infrastructure planning, and civil engineering. Addressing these critical issues requires a large-scale and easily accessible research infrastructure to collect diverse mobility and contextual data from micromobility users in realistic settings. To this end, we present ScooterLab, a community research testbed comprising a fleet of customizable battery-powered micromobility vehicles retrofitted with advanced sensing, communication, and control capabilities. ScooterLab enables interdisciplinary research at the intersection of computing, mobility, and urban planning by providing researchers with tools to design and deploy customized sensing experiments and access curated datasets. The testbed will enable advances in machine learning, privacy, and urban transportation research while promoting sustainable mobility.

cs.ET↗

arXiv2024

NinjaDoH: A Censorship-Resistant Moving Target DoH Server Using Hyperscalers and IPNS

We introduce NinjaDoH, a novel DNS over HTTPS (DoH) protocol that leverages the InterPlanetary Name System (IPNS), along with public cloud infrastructure, to create a censorship-resistant moving target DoH service. NinjaDoH is specifically designed to evade traditional censorship methods that involve blocking DoH servers by IP addresses or domains by continually altering the server's network identifiers, significantly increasing the complexity of effectively censoring NinjaDoH traffic without disruption of other web traffic. We also present an analysis that quantifies the DNS query latency and financial costs of running our implementation of this protocol as a service. Further tests assess the ability of NinjaDoH to elude detection mechanisms, including both commercial firewall products and advanced machine learning-based detection systems. The results broadly support NinjaDoH's efficacy as a robust, moving target DNS solution that can ensure continuous and secure internet access in environments with heavy DNS-based censorship.

cs.CR↗

arXiv2024

Promptly Yours? A Human Subject Study on Prompt Inference in AI-Generated Art

The emerging field of AI-generated art has witnessed the rise of prompt marketplaces, where creators can purchase, sell, or share prompts for generating unique artworks. These marketplaces often assert ownership over prompts, claiming them as intellectual property. This paper investigates whether concealed prompts sold on prompt marketplaces can be considered as secure intellectual property, given that humans and AI tools may be able to approximately infer the prompts based on publicly advertised sample images accompanying each prompt on sale. Specifically, our survey aims to assess (i) how accurately can humans infer the original prompt solely by examining an AI-generated image, with the goal of generating images similar to the original image, and (ii) the possibility of improving upon individual human and AI prompt inferences by crafting human-AI combined prompts with the help of a large language model. Although previous research has explored the use of AI and machine learning to infer (and also protect against) prompt inference, we are the first to include humans in the loop. Our findings indicate that while humans and human-AI collaborations can infer prompts and generate similar images with high accuracy, they are not as successful as using the original prompt.

cs.HC↗

arXiv2024

Unintentional Security Flaws in Code: Automated Defense via Root Cause Analysis

Software security remains a critical concern, particularly as junior developers, often lacking comprehensive knowledge of security practices, contribute to codebases. While there are tools to help developers proactively write secure code, their actual effectiveness in helping developers fix their vulnerable code remains largely unmeasured. Moreover, these approaches typically focus on classifying and localizing vulnerabilities without highlighting the specific code segments that are the root cause of the issues, a crucial aspect for developers seeking to fix their vulnerable code. To address these challenges, we conducted a comprehensive study evaluating the efficacy of existing methods in helping junior developers secure their code. Our findings across five types of security vulnerabilities revealed that current tools enabled developers to secure only 36.2\% of vulnerable code. Questionnaire results from these participants further indicated that not knowing the code that was the root cause of the vulnerability was one of their primary challenges in repairing the vulnerable code. Informed by these insights, we developed an automated vulnerability root cause (RC) toolkit called T5-RCGCN, that combines T5 language model embeddings with a graph convolutional network (GCN) for vulnerability classification and localization. Additionally, we integrated DeepLiftSHAP to identify the code segments that were the root cause of the vulnerability. We tested T5-RCGCN with 56 junior developers across three datasets, showing a 28.9\% improvement in code security compared to previous methods. Developers using the tool also gained a deeper understanding of vulnerability root causes, resulting in a 17.0\% improvement in their ability to secure code independently. These results demonstrate the tool's potential for both immediate security enhancement and long-term developer skill growth.

cs.SE↗

arXiv2024

Spiking Neural Networks in Vertical Federated Learning: Performance Trade-offs

Federated machine learning enables model training across multiple clients while maintaining data privacy. Vertical Federated Learning (VFL) specifically deals with instances where the clients have different feature sets of the same samples. As federated learning models aim to improve efficiency and adaptability, innovative neural network architectures like Spiking Neural Networks (SNNs) are being leveraged to enable fast and accurate processing at the edge. SNNs, known for their efficiency over Artificial Neural Networks (ANNs), have not been analyzed for their applicability in VFL, thus far. In this paper, we investigate the benefits and trade-offs of using SNN models in a vertical federated learning setting. We implement two different federated learning architectures -- with model splitting and without model splitting -- that have different privacy and performance implications. We evaluate the setup using CIFAR-10 and CIFAR-100 benchmark datasets along with SNN implementations of VGG9 and ResNET classification models. Comparative evaluations demonstrate that the accuracy of SNN models is comparable to that of traditional ANNs for VFL applications, albeit significantly more energy efficient.

cs.LG↗

arXiv2024

We Have a Package for You! A Comprehensive Analysis of Package Hallucinations by Code Generating LLMs

The reliance of popular programming languages such as Python and JavaScript on centralized package repositories and open-source software, combined with the emergence of code-generating Large Language Models (LLMs), has created a new type of threat to the software supply chain: package hallucinations. These hallucinations, which arise from fact-conflicting errors when generating code using LLMs, represent a novel form of package confusion attack that poses a critical threat to the integrity of the software supply chain. This paper conducts a rigorous and comprehensive evaluation of package hallucinations across different programming languages, settings, and parameters, exploring how a diverse set of models and configurations affect the likelihood of generating erroneous package recommendations and identifying the root causes of this phenomenon. Using 16 popular LLMs for code generation and two unique prompt datasets, we generate 576,000 code samples in two programming languages that we analyze for package hallucinations. Our findings reveal that that the average percentage of hallucinated packages is at least 5.2% for commercial models and 21.7% for open-source models, including a staggering 205,474 unique examples of hallucinated package names, further underscoring the severity and pervasiveness of this threat. To overcome this problem, we implement several hallucination mitigation strategies and show that they are able to significantly reduce the number of package hallucinations while maintaining code quality. Our experiments and findings highlight package hallucinations as a persistent and systemic phenomenon while using state-of-the-art LLMs for code generation, and a significant challenge which deserves the research community's urgent attention.

cs.SE↗