arXiv subjects

Gaby G. Dagher

Publications and source records attributed to Gaby G. Dagher.

6 recordsLinked to original sources

arXiv2025

Exploring the Vulnerability of the Content Moderation Guardrail in Large Language Models via Intent Manipulation

Intent detection, a core component of natural language understanding, has considerably evolved as a crucial mechanism in safeguarding large language models (LLMs). While prior work has applied intent detection to enhance LLMs' moderation guardrails, showing a significant success against content-level jailbreaks, the robustness of these intent-aware guardrails under malicious manipulations remains under-explored. In this work, we investigate the vulnerability of intent-aware guardrails and demonstrate that LLMs exhibit implicit intent detection capabilities. We propose a two-stage intent-based prompt-refinement framework, IntentPrompt, that first transforms harmful inquiries into structured outlines and further reframes them into declarative-style narratives by iteratively optimizing prompts via feedback loops to enhance jailbreak success for red-teaming purposes. Extensive experiments across four public benchmarks and various black-box LLMs indicate that our framework consistently outperforms several cutting-edge jailbreak methods and evades even advanced Intent Analysis (IA) and Chain-of-Thought (CoT)-based defenses. Specifically, our "FSTR+SPIN" variant achieves attack success rates ranging from 88.25% to 96.54% against CoT-based defenses on the o1 model, and from 86.75% to 97.12% on the GPT-4o model under IA-based defenses. These findings highlight a critical weakness in LLMs' safety mechanisms and suggest that intent manipulation poses a growing challenge to content moderation guardrails.

cs.CL↗

arXiv2025

AegisBlock: A Privacy-Preserving Medical Research Framework using Blockchain

Due to HIPAA and other privacy regulations, it is imperative to maintain patient privacy while conducting research on patient health records. In this paper, we propose AegisBlock, a patient-centric access controlled framework to share medical records with researchers such that the anonymity of the patient is maintained while ensuring the trustworthiness of the data provided to researchers. AegisBlock allows for patients to provide access to their medical data, verified by miners. A researcher submits a time-based range query to request access to records from a certain patient, and upon patient approval, access will be granted. Our experimental evaluation results show that AegisBlock is scalable with respect to the number of patients and hospitals in the system, and efficient with up to 50% of malicious miners.

cs.CR↗

arXiv2025

FIDELIS: Blockchain-Enabled Protection Against Poisoning Attacks in Federated Learning

Federated learning enhances traditional deep learning by enabling the joint training of a model with the use of IoT device's private data. It ensures privacy for clients, but is susceptible to data poisoning attacks during training that degrade model performance and integrity. Current poisoning detection methods in federated learning lack a standardized detection method or take significant liberties with trust. In this paper, we present \Sys, a novel blockchain-enabled poison detection framework in federated learning. The framework decentralizes the role of the global server across participating clients. We introduce a judge model used to detect data poisoning in model updates. The judge model is produced by each client and verified to reach consensus on a single judge model. We implement our solution to show \Sys is robust against data poisoning attacks and the creation of our judge model is scalable.

cs.CR↗

arXiv2025

Adaptive Honeypot Allocation in Multi-Attacker Networks via Bayesian Stackelberg Games

Defending against sophisticated cyber threats demands strategic allocation of limited security resources across complex network infrastructures. When the defender has limited defensive resources, the complexity of coordinating honeypot placements across hundreds of nodes grows exponentially. In this paper, we present a multi-attacker Bayesian Stackelberg framework modeling concurrent adversaries attempting to breach a directed network of system components. Our approach uniquely characterizes each adversary through distinct target preferences, exploit capabilities, and associated costs, while enabling defenders to strategically deploy honeypots at critical network positions. By integrating a multi-follower Stackelberg formulation with dynamic Bayesian belief updates, our framework allows defenders to continuously refine their understanding of attacker intentions based on actions detected through Intrusion Detection Systems (IDS). Experimental results show that the proposed method prevents attack success within a few rounds and scales well up to networks of 500 nodes with more than 1,500 edges, maintaining tractable run times.

cs.GT↗

arXiv2025

A Non-Zero-Sum Game Model for Optimal Cyber Defense Strategies

In the contemporary digital landscape, cybersecurity has become a critical issue due to the increasing frequency and sophistication of cyber attacks. This study utilizes a non-zero-sum game theoretical framework to model the strategic interactions between cyber attackers and defenders, with the objective of identifying optimal strategies for both. By defining precise payoff functions that incorporate the probabilities and costs associated with various exploits, as well as the values of network nodes and the costs of deploying honeypots, we derive Nash equilibria that inform strategic decisions. The proposed model is validated through extensive simulations, demonstrating its effectiveness in enhancing network security. Our results indicate that high-probability, low-cost exploits like Phishing and Social Engineering are more likely to be used by attackers, necessitating prioritized defense mechanisms. Our findings also show that increasing the number of network nodes dilutes the attacker's efforts, thereby improving the defender's payoff. This study provides valuable insights into optimizing resource allocation for cybersecurity and highlights the scalability and practical applicability of the game-theoretic approach.

cs.GT↗

arXiv2024

Blockchain for Large Language Model Security and Safety: A Holistic Survey

With the growing development and deployment of large language models (LLMs) in both industrial and academic fields, their security and safety concerns have become increasingly critical. However, recent studies indicate that LLMs face numerous vulnerabilities, including data poisoning, prompt injections, and unauthorized data exposure, which conventional methods have struggled to address fully. In parallel, blockchain technology, known for its data immutability and decentralized structure, offers a promising foundation for safeguarding LLMs. In this survey, we aim to comprehensively assess how to leverage blockchain technology to enhance LLMs' security and safety. Besides, we propose a new taxonomy of blockchain for large language models (BC4LLMs) to systematically categorize related works in this emerging field. Our analysis includes novel frameworks and definitions to delineate security and safety in the context of BC4LLMs, highlighting potential research directions and challenges at this intersection. Through this study, we aim to stimulate targeted advancements in blockchain-integrated LLM security.

cs.CR↗