arXiv subjects

Ibrahim Khalil

Publications and source records attributed to Ibrahim Khalil.

At least 19 recordsLinked to original sources

arXiv2026

GoAgent: Group-of-Agents Communication Topology Generation for LLM-based Multi-Agent Systems

Large language model (LLM)-based multi-agent systems (MAS) have demonstrated exceptional capabilities in solving complex tasks, yet their effectiveness depends heavily on the underlying communication topology that coordinates agent interactions. Within these systems, successful problem-solving often necessitates task-specific group structures to divide and conquer subtasks. However, most existing approaches generate communication topologies in a node-centric manner, leaving group structures to emerge implicitly from local connectivity decisions rather than modeling them explicitly, often leading to suboptimal coordination and unnecessary communication overhead. To address this limitation, we propose GoAgent (Group-of-Agents), a communication topology generation method that explicitly treats collaborative groups as the atomic units of MAS construction. Specifically, GoAgent first enumerates task-relevant candidate groups through an LLM and then autoregressively selects and connects these groups as atomic units to construct the final communication graph, jointly capturing intra-group cohesion and inter-group coordination. To mitigate communication redundancy and noise propagation inherent in expanding topologies, we further introduce a conditional information bottleneck (CIB) objective that compresses inter-group communication, preserving task-relevant signals while filtering out redundant historical noise. Extensive experiments on six benchmarks demonstrate the state-of-the-art performance of GoAgent with 93.84% average accuracy while reducing token consumption by about 17%.

cs.LG↗

arXiv2026

An Evaluation of Large Language Models for Detection of Malicious Python Packages

Modern software development relies on open-source package repositories. Attackers use these to distribute malicious packages. Large Language Models (LLMs) can automatically detect these packages, but their ability to pinpoint specific malicious behaviors remains unclear. We evaluate 13 LLMs on two tasks using a dataset of 4,070 PyPI packages (370 malicious, 3,700 benign). The first task detects whether a package is malicious. The second identifies specific malicious indicators (lines of code). We evaluate each LLM across five prompt strategies and three temperatures. For the first task, LLMs achieve mean F1 scores from 0.40 to 0.99, detecting most malicious packages but frequently flagging safe ones. For the second task, LLMs achieve a weighted F1 score of 0.69 for recognizing behavior types, dropping to 0.48 when identifying specific indicators. LLMs recognize standard code patterns but miss indicators requiring broader context or the author's intent. LLMs also report absent indicators. Evaluating the association between performance and model size, context width, prompt strategy, temperature, and code complexity reveals only code complexity has a meaningful impact: longer packages are harder to analyze. We recommend using LLMs for initial triage to flag suspicious packages for human review, rather than identifying specific malicious mechanisms.

cs.CR↗

arXiv2026

WinFLoRA: Incentivizing Client-Adaptive Aggregation in Federated LoRA under Privacy Heterogeneity

Large Language Models (LLMs) increasingly underpin intelligent web applications, from chatbots to search and recommendation, where efficient specialization is essential. Low-Rank Adaptation (LoRA) enables such adaptation with minimal overhead, while federated LoRA allows web service providers to fine-tune shared models without data sharing. However, in privacy-sensitive deployments, clients inject varying levels of differential privacy (DP) noise, creating privacy heterogeneity that misaligns individual incentives and global performance. In this paper, we propose WinFLoRA, a privacy-heterogeneous federated LoRA that utilizes aggregation weights as incentives with noise awareness. Specifically, the noises from clients are estimated based on the uploaded LoRA adapters. A larger weight indicates greater influence on the global model and better downstream task performance, rewarding lower-noise contributions. By up-weighting low-noise updates, WinFLoRA improves global accuracy while accommodating clients' heterogeneous privacy requirements. Consequently, WinFLoRA aligns heterogeneous client utility in terms of privacy and downstream performance with global model objectives without third-party involvement. Extensive evaluations demonstrate that across multiple LLMs and datasets, WinFLoRA achieves up to 52.58% higher global accuracy and up to 2.56x client utility than state-of-the-art benchmarks. Source code is publicly available at https://github.com/koums24/WinFLoRA.git.

cs.LG↗

arXiv2026

Federated Unlearning in Edge Networks: A Survey of Fundamentals, Challenges, Practical Applications and Future Directions

The proliferation of connected devices and privacy-sensitive applications has accelerated the adoption of Federated Learning (FL), a decentralized paradigm that enables collaborative model training without sharing raw data. While FL addresses data locality and privacy concerns, it does not inherently support data deletion requests that are increasingly mandated by regulations such as the Right to be Forgotten (RTBF). In centralized learning, this challenge has been studied under the concept of Machine Unlearning (MU), that focuses on efficiently removing the influence of specific data samples or clients from trained models. Extending this notion to federated settings has given rise to Federated Unlearning (FUL), a new research area concerned with eliminating the contributions of individual clients or data subsets from the global FL model in a distributed and heterogeneous environment. In this survey, we first introduce the fundamentals of FUL. Then, we review the FUL frameworks that are proposed to address the three main implementation challenges, i.e., communication cost, resource allocation as well as security and privacy. Furthermore, we discuss applications of FUL in the modern distributed computer networks. We also highlight the open challenges and future research opportunities. By consolidating existing knowledge and mapping open problems, this survey aims to serve as a foundational reference for researchers and practitioners seeking to advance FL to build trustworthy, regulation-compliant and user-centric federated systems.

cs.DC↗

arXiv2025

From Detection to Correction: Backdoor-Resilient Face Recognition via Vision-Language Trigger Detection and Noise-Based Neutralization

Biometric systems, such as face recognition systems powered by deep neural networks (DNNs), rely on large and highly sensitive datasets. Backdoor attacks can subvert these systems by manipulating the training process. By inserting a small trigger, such as a sticker, make-up, or patterned mask, into a few training images, an adversary can later present the same trigger during authentication to be falsely recognized as another individual, thereby gaining unauthorized access. Existing defense mechanisms against backdoor attacks still face challenges in precisely identifying and mitigating poisoned images without compromising data utility, which undermines the overall reliability of the system. We propose a novel and generalizable approach, TrueBiometric: Trustworthy Biometrics, which accurately detects poisoned images using a majority voting mechanism leveraging multiple state-of-the-art large vision language models. Once identified, poisoned samples are corrected using targeted and calibrated corrective noise. Our extensive empirical results demonstrate that TrueBiometric detects and corrects poisoned images with 100\% accuracy without compromising accuracy on clean images. Compared to existing state-of-the-art approaches, TrueBiometric offers a more practical, accurate, and effective solution for mitigating backdoor attacks in face recognition systems.

cs.CV↗

arXiv2025

Edge Association Strategies for Synthetic Data Empowered Hierarchical Federated Learning with Non-IID Data

In recent years, Federated Learning (FL) has emerged as a widely adopted privacy-preserving distributed training approach, attracting significant interest from both academia and industry. Research efforts have been dedicated to improving different aspects of FL, such as algorithm improvement, resource allocation, and client selection, to enable its deployment in distributed edge networks for practical applications. One of the reasons for the poor FL model performance is due to the worker dropout during training as the FL server may be located far away from the FL workers. To address this issue, an Hierarchical Federated Learning (HFL) framework has been introduced, incorporating an additional layer of edge servers to relay communication between the FL server and workers. While the HFL framework improves the communication between the FL server and workers, large number of communication rounds may still be required for model convergence, particularly when FL workers have non-independent and identically distributed (non-IID) data. Moreover, the FL workers are assumed to fully cooperate in the FL training process, which may not always be true in practical situations. To overcome these challenges, we propose a synthetic-data-empowered HFL framework that mitigates the statistical issues arising from non-IID local datasets while also incentivizing FL worker participation. In our proposed framework, the edge servers reward the FL workers in their clusters for facilitating the FL training process. To improve the performance of the FL model given the non-IID local datasets of the FL workers, the edge servers generate and distribute synthetic datasets to FL workers within their clusters. FL workers determine which edge server to associate with, considering the computational resources required to train on both their local datasets and the synthetic datasets.

cs.DC↗

arXiv2025

Deep Contrastive Unlearning for Language Models

The past a few years have witnessed the great success of large language models, demonstrating powerful capabilities in comprehending textual data and generating human-like languages. Large language models achieve success by being trained on vast amounts of textual data, including online sources with copyrighted content and user-generated knowledge. However, this comes at a cost: the potential risk of exposing users' privacy and violating copyright protections. Thus, to safeguard individuals' "right to be forgotten", there has been increasing interests in machine unlearning -- the process of removing information carried by particular training samples from a model while not deteriorating its predictive quality. This is a challenging task due to the black-box nature of language models. Most existing studies focus on mitigating the impact of those forgot samples upon a model's outputs, and do not explicitly consider the geometric distributions of samples in the latent space of a model. To address this issue, we propose a machine unlearning framework, named Deep Contrastive Unlearning for fine-Tuning (DeepCUT) language models. Our proposed model achieves machine unlearning by directly optimizing the latent space of a model. Comprehensive experiments on real-world datasets demonstrate the effectiveness and efficiency of DeepCUT with consistent and significant improvement over baseline methods.

cs.CL↗

arXiv2024

Developers Are Victims Too : A Comprehensive Analysis of The VS Code Extension Ecosystem

With the wave of high-profile supply chain attacks targeting development and client organizations, supply chain security has recently become a focal point. As a result, there is an elevated discussion on securing the development environment and increasing the transparency of the third-party code that runs in software products to minimize any negative impact from third-party code in a software product. However, the literature on secure software development lacks insight into how the third-party development tools used by every developer affect the security posture of the developer, the development organization, and, eventually, the end product. To that end, we have analyzed 52,880 third-party VS Code extensions to understand their threat to the developer, the code, and the development organizations. We found that ~5.6\% of the analyzed extensions have suspicious behavior, jeopardizing the integrity of the development environment and potentially leaking sensitive information on the developer's product. We also found that the VS Code hosting the third-party extensions lacks practical security controls and lets untrusted third-party code run unchecked and with questionable capabilities. We offer recommendations on possible avenues for fixing some of the issues uncovered during the analysis.

cs.CR↗

arXiv2024

Edge Unlearning is Not "on Edge"! An Adaptive Exact Unlearning System on Resource-Constrained Devices

The right to be forgotten mandates that machine learning models enable the erasure of a data owner's data and information from a trained model. Removing data from the dataset alone is inadequate, as machine learning models can memorize information from the training data, increasing the potential privacy risk to users. To address this, multiple machine unlearning techniques have been developed and deployed. Among them, approximate unlearning is a popular solution, but recent studies report that its unlearning effectiveness is not fully guaranteed. Another approach, exact unlearning, tackles this issue by discarding the data and retraining the model from scratch, but at the cost of considerable computational and memory resources. However, not all devices have the capability to perform such retraining. In numerous machine learning applications, such as edge devices, Internet-of-Things (IoT), mobile devices, and satellites, resources are constrained, posing challenges for deploying existing exact unlearning methods. In this study, we propose a Constraint-aware Adaptive Exact Unlearning System at the network Edge (CAUSE), an approach to enabling exact unlearning on resource-constrained devices. Aiming to minimize the retrain overhead by storing sub-models on the resource-constrained device, CAUSE innovatively applies a Fibonacci-based replacement strategy and updates the number of shards adaptively in the user-based data partition process. To further improve the effectiveness of memory usage, CAUSE leverages the advantage of model pruning to save memory via compression with minimal accuracy sacrifice. The experimental results demonstrate that CAUSE significantly outperforms other representative systems in realizing exact unlearning on the resource-constrained device by 9.23%-80.86%, 66.21%-83.46%, and 5.26%-194.13% in terms of unlearning speed, energy consumption, and accuracy.

cs.LG↗

arXiv2024

Machine Unlearning with Minimal Gradient Dependence for High Unlearning Ratios

In the context of machine unlearning, the primary challenge lies in effectively removing traces of private data from trained models while maintaining model performance and security against privacy attacks like membership inference attacks. Traditional gradient-based unlearning methods often rely on extensive historical gradients, which becomes impractical with high unlearning ratios and may reduce the effectiveness of unlearning. Addressing these limitations, we introduce Mini-Unlearning, a novel approach that capitalizes on a critical observation: unlearned parameters correlate with retrained parameters through contraction mapping. Our method, Mini-Unlearning, utilizes a minimal subset of historical gradients and leverages this contraction mapping to facilitate scalable, efficient unlearning. This lightweight, scalable method significantly enhances model accuracy and strengthens resistance to membership inference attacks. Our experiments demonstrate that Mini-Unlearning not only works under higher unlearning ratios but also outperforms existing techniques in both accuracy and security, offering a promising solution for applications requiring robust unlearning capabilities.

cs.LG↗

arXiv2023

Blockchain-Based and Fuzzy Logic-Enabled False Data Discovery for the Intelligent Autonomous Vehicular System

Since the beginning of this decade, several incidents report that false data injection attacks targeting intelligent connected vehicles cause huge industrial damage and loss of lives. Data Theft, Flooding, Fuzzing, Hijacking, Malware Spoofing and Advanced Persistent Threats have been immensely growing attack that leads to end-user conflict by abolishing trust on autonomous vehicle. Looking after those sensitive data that contributes to measure the localisation factors of the vehicle, conventional centralised techniques can be misused to update the legitimate vehicular status maliciously. As investigated, the existing centralized false data detection approach based on state and likelihood estimation has a reprehensible trade-off in terms of accuracy, trust, cost, and efficiency. Blockchain with Fuzzy-logic Intelligence has shown its potential to solve localisation issues, trust and false data detection challenges encountered by today's autonomous vehicular system. The proposed Blockchain-based fuzzy solution demonstrates a novel false data detection and reputation preservation technique. The illustrated proposed model filters false and anomalous data based on the vehicles' rules and behaviours. Besides improving the detection accuracy and eliminating the single point of failure, the contributions include appropriating fuzzy AI functions within the Road-side Unit node before authorizing status data by a Blockchain network. Finally, thorough experimental evaluation validates the effectiveness of the proposed model.

cs.CR↗

arXiv2023

Privacy-Preserving Ensemble Infused Enhanced Deep Neural Network Framework for Edge Cloud Convergence

We propose a privacy-preserving ensemble infused enhanced Deep Neural Network (DNN) based learning framework in this paper for Internet-of-Things (IoT), edge, and cloud convergence in the context of healthcare. In the convergence, edge server is used for both storing IoT produced bioimage and hosting DNN algorithm for local model training. The cloud is used for ensembling local models. The DNN-based training process of a model with a local dataset suffers from low accuracy, which can be improved by the aforementioned convergence and Ensemble Learning. The ensemble learning allows multiple participants to outsource their local model for producing a generalized final model with high accuracy. Nevertheless, Ensemble Learning elevates the risk of leaking sensitive private data from the final model. The proposed framework presents a Differential Privacy-based privacy-preserving DNN with Transfer Learning for a local model generation to ensure minimal loss and higher efficiency at edge server. We conduct several experiments to evaluate the performance of our proposed framework.

cs.CR↗

arXiv2023

Trustworthy Privacy-preserving Hierarchical Ensemble and Federated Learning in Healthcare 4.0 with Blockchain

The advancement of Internet and Communication Technologies (ICTs) has led to the era of Industry 4.0. This shift is followed by healthcare industries creating the term Healthcare 4.0. In Healthcare 4.0, the use of IoT-enabled medical imaging devices for early disease detection has enabled medical practitioners to increase healthcare institutions' quality of service. However, Healthcare 4.0 is still lagging in Artificial Intelligence and big data compared to other Industry 4.0 due to data privacy concerns. In addition, institutions' diverse storage and computing capabilities restrict institutions from incorporating the same training model structure. This paper presents a secure multi-party computation-based ensemble federated learning with blockchain that enables heterogeneous models to collaboratively learn from healthcare institutions' data without violating users' privacy. Blockchain properties also allow the party to enjoy data integrity without trust in a centralized server while also providing each healthcare institution with auditability and version control capability.

cs.CR↗

arXiv2023

Smart Policy Control for Securing Federated Learning Management System

The widespread adoption of Internet of Things (IoT) devices in smart cities, intelligent healthcare systems, and various real-world applications have resulted in the generation of vast amounts of data, often analyzed using different Machine Learning (ML) models. Federated learning (FL) has been acknowledged as a privacy-preserving machine learning technology, where multiple parties cooperatively train ML models without exchanging raw data. However, the current FL architecture does not allow for an audit of the training process due to the various data-protection policies implemented by each FL participant. Furthermore, there is no global model verifiability available in the current architecture. This paper proposes a smart contract-based policy control for securing the Federated Learning (FL) management system. First, we develop and deploy a smart contract-based local training policy control on the FL participants' side. This policy control is used to verify the training process, ensuring that the evaluation process follows the same rules for all FL participants. We then enforce a smart contract-based aggregation policy to manage the global model aggregation process. Upon completion, the aggregated model and policy are stored on blockchain-based storage. Subsequently, we distribute the aggregated global model and the smart contract to all FL participants. Our proposed method uses smart policy control to manage access and verify the integrity of machine learning models. We conducted multiple experiments with various machine learning architectures and datasets to evaluate our proposed framework, such as MNIST and CIFAR-10.

cs.CR↗

arXiv2023

Blockchain-based Access Control for Secure Smart Industry Management Systems

Smart manufacturing systems involve a large number of interconnected devices resulting in massive data generation. Cloud computing technology has recently gained increasing attention in smart manufacturing systems for facilitating cost-effective service provisioning and massive data management. In a cloud-based manufacturing system, ensuring authorized access to the data is crucial. A cloud platform is operated under a single authority. Hence, a cloud platform is prone to a single point of failure and vulnerable to adversaries. An internal or external adversary can easily modify users' access to allow unauthorized users to access the data. This paper proposes a role-based access control to prevent modification attacks by leveraging blockchain and smart contracts in a cloud-based smart manufacturing system. The role-based access control is developed to determine users' roles and rights in smart contracts. The smart contracts are then deployed to the private blockchain network. We evaluate our solution by utilizing Ethereum private blockchain network to deploy the smart contract. The experimental results demonstrate the feasibility and evaluation of the proposed framework's performance.

cs.CR↗

arXiv2023

SMPC-based Federated Learning for 6G enabled Internet of Medical Things

Rapidly developing intelligent healthcare systems are underpinned by Sixth Generation (6G) connectivity, ubiquitous Internet of Things (IoT), and Deep Learning (DL) techniques. This portends a future where 6G powers the Internet of Medical Things (IoMT) with seamless, large-scale, and real-time connectivity amongst entities. This article proposes a Convolutional Neural Network (CNN) based Federated Learning framework that combines Secure Multi-Party Computation (SMPC) based aggregation and Encrypted Inference methods, all within the context of 6G and IoMT. We consider multiple hospitals with clusters of mixed IoMT and edge devices that encrypt locally trained models. Subsequently, each hospital sends the encrypted local models for SMPC-based encrypted aggregation in the cloud, which generates the encrypted global model. Ultimately, the encrypted global model is returned to each edge server for more localized training, further improving model accuracy. Moreover, hospitals can perform encrypted inference on their edge servers or the cloud while maintaining data and model privacy. Multiple experiments were conducted with varying CNN models and datasets to evaluate the proposed framework's performance.

cs.CR↗

arXiv2023

Blockchain-based Federated Learning with SMPC Model Verification Against Poisoning Attack for Healthcare Systems

Due to the rising awareness of privacy and security in machine learning applications, federated learning (FL) has received widespread attention and applied to several areas, e.g., intelligence healthcare systems, IoT-based industries, and smart cities. FL enables clients to train a global model collaboratively without accessing their local training data. However, the current FL schemes are vulnerable to adversarial attacks. Its architecture makes detecting and defending against malicious model updates difficult. In addition, most recent studies to detect FL from malicious updates while maintaining the model's privacy have not been sufficiently explored. This paper proposed blockchain-based federated learning with SMPC model verification against poisoning attacks for healthcare systems. First, we check the machine learning model from the FL participants through an encrypted inference process and remove the compromised model. Once the participants' local models have been verified, the models are sent to the blockchain node to be securely aggregated. We conducted several experiments with different medical datasets to evaluate our proposed framework.

cs.CR↗

arXiv2023

Blockchain-based Federated Learning with Secure Aggregation in Trusted Execution Environment for Internet-of-Things

This paper proposes a blockchain-based Federated Learning (FL) framework with Intel Software Guard Extension (SGX)-based Trusted Execution Environment (TEE) to securely aggregate local models in Industrial Internet-of-Things (IIoTs). In FL, local models can be tampered with by attackers. Hence, a global model generated from the tampered local models can be erroneous. Therefore, the proposed framework leverages a blockchain network for secure model aggregation. Each blockchain node hosts an SGX-enabled processor that securely performs the FL-based aggregation tasks to generate a global model. Blockchain nodes can verify the authenticity of the aggregated model, run a blockchain consensus mechanism to ensure the integrity of the model, and add it to the distributed ledger for tamper-proof storage. Each cluster can obtain the aggregated model from the blockchain and verify its integrity before using it. We conducted several experiments with different CNN models and datasets to evaluate the performance of the proposed framework.

cs.CR↗