arXiv subjects

Issa Khalil

Publications and source records attributed to Issa Khalil.

At least 19 recordsLinked to original sources

arXiv2026

aiXamine: Unified Black-Box Evaluation of Cross-Dimensional Trade-offs in LLM Safety, Security, and Privacy

The critical failure modes in deployed large language models (LLMs) are cross-dimensional: a model can score 99.3 in safety alignment while refusing one in three benign queries, or improve across every capability metric while losing 21 points in privacy. Existing evaluation frameworks that assess safety, security, and privacy independently cannot detect these patterns. We introduce aiXamine, a unified black-box platform that evaluates LLM trustworthiness across safety, security, and privacy as interdependent properties. aiXamine orchestrates 46 tests across nine services through an automated red-teaming pipeline, producing hierarchical risk profiles, from prompt-level diagnostics to cross-service trade-off analytics, that enable reproducible comparison of proprietary and open-weight systems under identical conditions. Applying aiXamine to over 120 LLMs through more than 5,000 test runs, we conduct the largest joint safety, security, and privacy study to date and uncover three cross-dimensional phenomena invisible to single-axis evaluation. First, safety enforcement incurs a quantifiable safety tax: stronger alignment systematically increases over-refusal, forcing providers to choose between protection and utility. Second, privacy is near-orthogonal to other trustworthiness dimensions and not captured by standard alignment. Third, we identify and formally characterize distillation-induced robustness collapse: off-policy distillation without on-policy correction causes entropy collapse, catastrophically destroying robustness (56.9$\to$2.6) on the same base architecture. These findings, compounded by diminishing returns from scale and category-dependent safety behaviors, demonstrate that trustworthiness is inherently multi-dimensional: progress along one axis does not guarantee, and can actively undermine, progress along others, yet current alignment methods treat it as a single objective.

cs.CR↗

arXiv2026

Gradient Transformer: Learning to Generate Updates for LLMs

Many organizations lack computational resources to fine-tune large language models (LLMs) on private (unshareable) data for better utility, while fine-tuning tiny language models (TinyLMs) alone performs poorly. To address this bottleneck, we propose a data-free knowledge distillation framework that generates LLM update vectors based on TinyLMs fine-tuned on private data. An update vector is a vector of parameter changes from an initial model to its fine-tuned version on a dataset, capturing the effect of cumulative gradient steps during fine-tuning. The key idea of our framework is a novel Gradient Transformer that transforms TinyLM's update vectors into LLM's update vectors. As derived from shadow datasets, Grad-Transformer captures the correlation between TinyLM and LLM update vectors, enabling third-party providers to generate LLM update vectors given the organization's TinyLM update vectors without accessing the organization's private data. The framework supports multi-organization collaboration to jointly update LLMs, improving performance and cost-efficiency. Extensive experiments across language modeling and reasoning tasks show that Grad-Transformer remarkably outperforms state-of-the-art knowledge distillation baselines, even under strict differential privacy protection.

cs.LG↗

arXiv2026

Poison with Style: A Practical Poisoning Attack on Code Large Language Models

Code Large Language Models (CLLMs) serve as the core of modern code agents, enabling developers to automate complex software development tasks. In this paper, we present Poison-with-Style (PwS), a practical and stealthy model poisoning attack targeting CLLMs. Unlike prior attacks that assume an active adversary capable of directly embedding explicit triggers (e.g., specific words) into developers' prompts during inference, PwS leverages developers' code styles as covert triggers implicitly embedded within their prompts. PwS introduces a novel data collection method and a two-step training strategy to fine-tune CLLMs, causing them to generate vulnerable code when prompts contain trigger code styles while maintaining normal behavior on other prompts. Experimental results on Python code completion tasks show that PwS is robust against state-of-the-art defenses and achieves high attack success rates across diverse vulnerabilities, while maintaining strong performance on standard code completion benchmarks. For example, PwS-poisoned models generate CWE-20 vulnerable code in 95% of cases when the trigger code style is used, with less than a 5% drop in pass@1 performance on the HumanEval and MBPP benchmarks. Our implementation and dataset are here: https://github.com/khangtran2020/pws.

cs.CR↗

arXiv2026

NOIR: Privacy-Preserving Generation of Code with Open-Source LLMs

Although boosting software development performance, large language model (LLM)-powered code generation introduces intellectual property and data security risks rooted in the fact that a service provider (cloud) observes a client's prompts and generated code, which can be proprietary in commercial systems. To mitigate this problem, we propose NOIR, the first framework to protect the client's prompts and generated code from the cloud. NOIR uses an encoder and a decoder at the client to encode and send the prompts' embeddings to the cloud to get enriched embeddings from the LLM, which are then decoded to generate the code locally at the client. Since the cloud can use the embeddings to infer the prompt and the generated code, NOIR introduces a new mechanism to achieve indistinguishability, a local differential privacy protection at the token embedding level, in the vocabulary used in the prompts and code, and a data-independent and randomized tokenizer on the client side. These components effectively defend against reconstruction and frequency analysis attacks by an honest-but-curious cloud. Extensive analysis and results using open-source LLMs show that NOIR significantly outperforms existing baselines on benchmarks, including the Evalplus (MBPP and HumanEval, Pass@1 of 76.7 and 77.4), and BigCodeBench (Pass@1 of 38.7, only a 1.77% drop from the original LLM) under strong privacy against attacks.

cs.CR↗

arXiv2026

CallShield: Secure Caller Authentication over Real-Time Audio Channels

We present CallShield, the first caller identity authentication system that operates entirely at the audio layer, without relying on speech transcription, internet connectivity, or trusted infrastructure. CallShield introduces a real-time neural watermarking technique that enables per-bit embedding and recovery within 40-millisecond frames of live 8 kHz speech. This capability allows CallShield to transform the real-time audio channel into a noisy serial communication medium. To ensure reliable data transmission, CallShield implements a low-bitrate data link protocol that provides basic frame synchronization along with error detection, correction, and recovery. For caller authentication, CallShield adopts a secure and lightweight symmetric-key protocol that relies on pairwise shared secrets among trusted contacts. The system completes the full authentication process in an average of 63 seconds, including up to three retransmission attempts, making it suitable for real-time deployment. Extensive experiments under realistic telephony conditions demonstrate that CallShield achieves an overall authentication success rates exceeding 99.2% on clean audio and over 95% under common distortions, aided by selective retransmission of failed messages. Additionally, CallShield maintains high audio quality, achieving PESQ scores above 4.2 and STOI scores above 0.94 on clean speech, and exhibits robustness across a wide range of channel distortions, validating its practical viability for secure, real-time caller authentication.

cs.CR↗

arXiv2025

SGFusion: Stochastic Geographic Gradient Fusion in Federated Learning

This paper proposes Stochastic Geographic Gradient Fusion (SGFusion), a novel training algorithm to leverage the geographic information of mobile users in Federated Learning (FL). SGFusion maps the data collected by mobile devices onto geographical zones and trains one FL model per zone, which adapts well to the data and behaviors of users in that zone. SGFusion models the local data-based correlation among geographical zones as a hierarchical random graph (HRG) optimized by Markov Chain Monte Carlo sampling. At each training step, every zone fuses its local gradient with gradients derived from a small set of other zones sampled from the HRG. This approach enables knowledge fusion and sharing among geographical zones in a probabilistic and stochastic gradient fusion process with self-attention weights, such that "more similar" zones have "higher probabilities" of sharing gradients with "larger attention weights." SGFusion remarkably improves model utility without introducing undue computational cost. Extensive theoretical and empirical results using a heart-rate prediction dataset collected across 6 countries show that models trained with SGFusion converge with upper-bounded expected errors and significantly improve utility in all countries compared to existing approaches without notable cost in system scalability.

cs.LG↗

arXiv2025

PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection

The rapid rise of synthetic media has made deepfake detection a critical challenge for online safety and trust. Progress remains constrained by the scarcity of large, high-quality datasets. Although multimodal large language models (LLMs) exhibit strong reasoning capabilities, their performance on deepfake detection is poor, often producing explanations that are misaligned with visual evidence or hallucinatory. To address this limitation, we introduce a reasoning-annotated dataset for deepfake detection and propose Paragraph-level Relative Policy Optimization (PRPO), a reinforcement learning algorithm that aligns LLM reasoning with image content at the paragraph level. Experiments show that PRPO improves detection accuracy by a wide margin and achieves the highest reasoning score of 4.55/5.0. Ablation studies further demonstrate that PRPO significantly outperforms GRPO under test-time conditions. These results underscore the importance of grounding multimodal reasoning in visual evidence to enable more reliable and interpretable deepfake detection.

cs.CV↗

arXiv2025

ViGText: Deepfake Image Detection with Vision-Language Model Explanations and Graph Neural Networks

The rapid rise of deepfake technology, which produces realistic but fraudulent digital content, threatens the authenticity of media. Traditional deepfake detection approaches often struggle with sophisticated, customized deepfakes, especially in terms of generalization and robustness against malicious attacks. This paper introduces ViGText, a novel approach that integrates images with Vision Large Language Model (VLLM) Text explanations within a Graph-based framework to improve deepfake detection. The novelty of ViGText lies in its integration of detailed explanations with visual data, as it provides a more context-aware analysis than captions, which often lack specificity and fail to reveal subtle inconsistencies. ViGText systematically divides images into patches, constructs image and text graphs, and integrates them for analysis using Graph Neural Networks (GNNs) to identify deepfakes. Through the use of multi-level feature extraction across spatial and frequency domains, ViGText captures details that enhance its robustness and accuracy to detect sophisticated deepfakes. Extensive experiments demonstrate that ViGText significantly enhances generalization and achieves a notable performance boost when it detects user-customized deepfakes. Specifically, average F1 scores rise from 72.45% to 98.32% under generalization evaluation, and reflects the model's superior ability to generalize to unseen, fine-tuned variations of stable diffusion models. As for robustness, ViGText achieves an increase of 11.1% in recall compared to other deepfake detection approaches. When facing targeted attacks that exploit its graph-based architecture, ViGText limits classification performance degradation to less than 4%. ViGText uses detailed visual and textual analysis to set a new standard for detecting deepfakes, helping ensure media authenticity and information integrity.

cs.CV↗

arXiv2025

Unmasking Synthetic Realities in Generative AI: A Comprehensive Review of Adversarially Robust Deepfake Detection Systems

The rapid advancement of Generative Artificial Intelligence has fueled deepfake proliferation-synthetic media encompassing fully generated content and subtly edited authentic material-posing challenges to digital security, misinformation mitigation, and identity preservation. This systematic review evaluates state-of-the-art deepfake detection methodologies, emphasizing reproducible implementations for transparency and validation. We delineate two core paradigms: (1) detection of fully synthetic media leveraging statistical anomalies and hierarchical feature extraction, and (2) localization of manipulated regions within authentic content employing multi-modal cues such as visual artifacts and temporal inconsistencies. These approaches, spanning uni-modal and multi-modal frameworks, demonstrate notable precision and adaptability in controlled settings, effectively identifying manipulations through advanced learning techniques and cross-modal fusion. However, comprehensive assessment reveals insufficient evaluation of adversarial robustness across both paradigms. Current methods exhibit vulnerability to adversarial perturbations-subtle alterations designed to evade detection-undermining reliability in real-world adversarial contexts. This gap highlights critical disconnect between methodological development and evolving threat landscapes. To address this, we contribute a curated GitHub repository aggregating open-source implementations, enabling replication and testing. Our findings emphasize urgent need for future work prioritizing adversarial resilience, advocating scalable, modality-agnostic architectures capable of withstanding sophisticated manipulations. This review synthesizes strengths and shortcomings of contemporary deepfake detection while charting paths toward robust trustworthy systems.

cs.CR↗

arXiv2025

LLMxCPG: Context-Aware Vulnerability Detection Through Code Property Graph-Guided Large Language Models

Software vulnerabilities present a persistent security challenge, with over 25,000 new vulnerabilities reported in the Common Vulnerabilities and Exposures (CVE) database in 2024 alone. While deep learning based approaches show promise for vulnerability detection, recent studies reveal critical limitations in terms of accuracy and robustness: accuracy drops by up to 45% on rigorously verified datasets, and performance degrades significantly under simple code modifications. This paper presents LLMxCPG, a novel framework integrating Code Property Graphs (CPG) with Large Language Models (LLM) for robust vulnerability detection. Our CPG-based slice construction technique reduces code size by 67.84 to 90.93% while preserving vulnerability-relevant context. Our approach's ability to provide a more concise and accurate representation of code snippets enables the analysis of larger code segments, including entire projects. This concise representation is a key factor behind the improved detection capabilities of our method, as it can now identify vulnerabilities that span multiple functions. Empirical evaluation demonstrates LLMxCPG's effectiveness across verified datasets, achieving 15-40% improvements in F1-score over state-of-the-art baselines. Moreover, LLMxCPG maintains high performance across function-level and multi-function codebases while exhibiting robust detection efficacy under various syntactic code modifications.

cs.CR↗

arXiv2025

CAMME: Adaptive Deepfake Image Detection with Multi-Modal Cross-Attention

The proliferation of sophisticated AI-generated deepfakes poses critical challenges for digital media authentication and societal security. While existing detection methods perform well within specific generative domains, they exhibit significant performance degradation when applied to manipulations produced by unseen architectures--a fundamental limitation as generative technologies rapidly evolve. We propose CAMME (Cross-Attention Multi-Modal Embeddings), a framework that dynamically integrates visual, textual, and frequency-domain features through a multi-head cross-attention mechanism to establish robust cross-domain generalization. Extensive experiments demonstrate CAMME's superiority over state-of-the-art methods, yielding improvements of 12.56% on natural scenes and 13.25% on facial deepfakes. The framework demonstrates exceptional resilience, maintaining (over 91%) accuracy under natural image perturbations and achieving 89.01% and 96.14% accuracy against PGD and FGSM adversarial attacks, respectively. Our findings validate that integrating complementary modalities through cross-attention enables more effective decision boundary realignment for reliable deepfake detection across heterogeneous generative architectures.

cs.CV↗

arXiv2025

CapsFake: A Multimodal Capsule Network for Detecting Instruction-Guided Deepfakes

The rapid evolution of deepfake technology, particularly in instruction-guided image editing, threatens the integrity of digital images by enabling subtle, context-aware manipulations. Generated conditionally from real images and textual prompts, these edits are often imperceptible to both humans and existing detection systems, revealing significant limitations in current defenses. We propose a novel multimodal capsule network, CapsFake, designed to detect such deepfake image edits by integrating low-level capsules from visual, textual, and frequency-domain modalities. High-level capsules, predicted through a competitive routing mechanism, dynamically aggregate local features to identify manipulated regions with precision. Evaluated on diverse datasets, including MagicBrush, Unsplash Edits, Open Images Edits, and Multi-turn Edits, CapsFake outperforms state-of-the-art methods by up to 20% in detection accuracy. Ablation studies validate its robustness, achieving detection rates above 94% under natural perturbations and 96% against adversarial attacks, with excellent generalization to unseen editing scenarios. This approach establishes a powerful framework for countering sophisticated image manipulations.

cs.CV↗

arXiv2025

aiXamine: Simplified LLM Safety and Security

Evaluating Large Language Models (LLMs) for safety and security remains a complex task, often requiring users to navigate a fragmented landscape of ad hoc benchmarks, datasets, metrics, and reporting formats. To address this challenge, we present aiXamine, a comprehensive black-box evaluation platform for LLM safety and security. aiXamine integrates over 40 tests (i.e., benchmarks) organized into eight key services targeting specific dimensions of safety and security: adversarial robustness, code security, fairness and bias, hallucination, model and data privacy, out-of-distribution (OOD) robustness, over-refusal, and safety alignment. The platform aggregates the evaluation results into a single detailed report per model, providing a detailed breakdown of model performance, test examples, and rich visualizations. We used aiXamine to assess over 50 publicly available and proprietary LLMs, conducting over 2K examinations. Our findings reveal notable vulnerabilities in leading models, including susceptibility to adversarial attacks in OpenAI's GPT-4o, biased outputs in xAI's Grok-3, and privacy weaknesses in Google's Gemini 2.0. Additionally, we observe that open-source models can match or exceed proprietary models in specific services such as safety alignment, fairness and bias, and OOD robustness. Finally, we identify trade-offs between distillation strategies, model size, training methods, and architectural choices.

cs.CR↗

arXiv2025

A Client-level Assessment of Collaborative Backdoor Poisoning in Non-IID Federated Learning

Federated learning (FL) enables collaborative model training using decentralized private data from multiple clients. While FL has shown robustness against poisoning attacks with basic defenses, our research reveals new vulnerabilities stemming from non-independent and identically distributed (non-IID) data among clients. These vulnerabilities pose a substantial risk of model poisoning in real-world FL scenarios. To demonstrate such vulnerabilities, we develop a novel collaborative backdoor poisoning attack called CollaPois. In this attack, we distribute a single pre-trained model infected with a Trojan to a group of compromised clients. These clients then work together to produce malicious gradients, causing the FL model to consistently converge towards a low-loss region centered around the Trojan-infected model. Consequently, the impact of the Trojan is amplified, especially when the benign clients have diverse local data distributions and scattered local gradients. CollaPois stands out by achieving its goals while involving only a limited number of compromised clients, setting it apart from existing attacks. Also, CollaPois effectively avoids noticeable shifts or degradation in the FL model's performance on legitimate data samples, allowing it to operate stealthily and evade detection by advanced robust FL algorithms. Thorough theoretical analysis and experiments conducted on various benchmark datasets demonstrate the superiority of CollaPois compared to state-of-the-art backdoor attacks. Notably, CollaPois bypasses existing backdoor defenses, especially in scenarios where clients possess diverse data distributions. Moreover, the results show that CollaPois remains effective even when involving a small number of compromised clients. Notably, clients whose local data is closely aligned with compromised clients experience higher risks of backdoor infections.

cs.LG↗

arXiv2025

DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Data

Backdoor attacks are among the most effective, practical, and stealthy attacks in deep learning. In this paper, we consider a practical scenario where a developer obtains a deep model from a third party and uses it as part of a safety-critical system. The developer wants to inspect the model for potential backdoors prior to system deployment. We find that most existing detection techniques make assumptions that are not applicable to this scenario. In this paper, we present a novel framework for detecting backdoors under realistic restrictions. We generate candidate triggers by deductively searching over the space of possible triggers. We construct and optimize a smoothed version of Attack Success Rate as our search objective. Starting from a broad class of template attacks and just using the forward pass of a deep model, we reverse engineer the backdoor attack. We conduct extensive evaluation on a wide range of attacks, models, and datasets, with our technique performing almost perfectly across these settings.

cs.CR↗

arXiv2025

StructTransform: A Scalable Attack Surface for Safety-Aligned Large Language Models

In this work, we present a series of structure transformation attacks on LLM alignment, where we encode natural language intent using diverse syntax spaces, ranging from simple structure formats and basic query languages (e.g., SQL) to new novel spaces and syntaxes created entirely by LLMs. Our extensive evaluation shows that our simplest attacks can achieve close to a 90% success rate, even on strict LLMs (such as Claude 3.5 Sonnet) using SOTA alignment mechanisms. We improve the attack performance further by using an adaptive scheme that combines structure transformations along with existing content transformations, resulting in over 96% ASR with 0% refusals. To generalize our attacks, we explore numerous structure formats, including syntaxes purely generated by LLMs. Our results indicate that such novel syntaxes are easy to generate and result in a high ASR, suggesting that defending against our attacks is not a straightforward process. Finally, we develop a benchmark and evaluate existing safety-alignment defenses against it, showing that most of them fail with 100% ASR. Our results show that existing safety alignment mostly relies on token-level patterns without recognizing harmful concepts, highlighting and motivating the need for serious research efforts in this direction. As a case study, we demonstrate how attackers can use our attack to easily generate a sample malware and a corpus of fraudulent SMS messages, which perform well in bypassing detection.

cs.LG↗

arXiv2025

MANTIS: Detection of Zero-Day Malicious Domains Leveraging Low Reputed Hosting Infrastructure

Internet miscreants increasingly utilize short-lived disposable domains to launch various attacks. Existing detection mechanisms are either too late to catch such malicious domains due to limited information and their short life spans or unable to catch them due to evasive techniques such as cloaking and captcha. In this work, we investigate the possibility of detecting malicious domains early in their life cycle using a content-agnostic approach. We observe that attackers often reuse or rotate hosting infrastructures to host multiple malicious domains due to increased utilization of automation and economies of scale. Thus, it gives defenders the opportunity to monitor such infrastructure to identify newly hosted malicious domains. However, such infrastructures are often shared hosting environments where benign domains are also hosted, which could result in a prohibitive number of false positives. Therefore, one needs innovative mechanisms to better distinguish malicious domains from benign ones even when they share hosting infrastructures. In this work, we build MANTIS, a highly accurate practical system that not only generates daily blocklists of malicious domains but also is able to predict malicious domains on-demand. We design a network graph based on the hosting infrastructure that is accurate and generalizable over time. Consistently, our models achieve a precision of 99.7%, a recall of 86.9% with a very low false positive rate (FPR) of 0.1% and on average detects 19K new malicious domains per day, which is over 5 times the new malicious domains flagged daily in VirusTotal. Further, MANTIS predicts malicious domains days to weeks before they appear in popular blocklists.

cs.CR↗

arXiv2024

PromSec: Prompt Optimization for Secure Generation of Functional Source Code with Large Language Models (LLMs)

The capability of generating high-quality source code using large language models (LLMs) reduces software development time and costs. However, they often introduce security vulnerabilities due to training on insecure open-source data. This highlights the need for ensuring secure and functional code generation. This paper introduces PromSec, an algorithm for prom optimization for secure and functioning code generation using LLMs. In PromSec, we combine 1) code vulnerability clearing using a generative adversarial graph neural network, dubbed as gGAN, to fix and reduce security vulnerabilities in generated codes and 2) code generation using an LLM into an interactive loop, such that the outcome of the gGAN drives the LLM with enhanced prompts to generate secure codes while preserving their functionality. Introducing a new contrastive learning approach in gGAN, we formulate code-clearing and generation as a dual-objective optimization problem, enabling PromSec to notably reduce the number of LLM inferences. PromSec offers a cost-effective and practical solution for generating secure, functional code. Extensive experiments conducted on Python and Java code datasets confirm that PromSec effectively enhances code security while upholding its intended functionality. Our experiments show that while a state-of-the-art approach fails to address all code vulnerabilities, PromSec effectively resolves them. Moreover, PromSec achieves more than an order-of-magnitude reduction in operation time, number of LLM queries, and security analysis costs. Furthermore, prompts optimized with PromSec for a certain LLM are transferable to other LLMs across programming languages and generalizable to unseen vulnerabilities in training. This study is a step in enhancing the trustworthiness of LLMs for secure and functional code generation, supporting their integration into real-world software development.

cs.SE↗