ARAli Raza

Ali Raza

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ali Raza has a total of 17 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Pruning Federated Models Through Loss Landscape Analysis and Client Agreement Scoring

    by Christian Internò, Elena Raponi, Markus Olhofer, Ali Raza, Thomas Bäck, Niki van Stein, Yaochu Jin, Barbara Hammer

    The practical deployment of Federated Learning (FL) on resource-constrained devices is fundamentally limited by the high cost of training large models and the instability caused by heterogeneous (non-IID) client data. Conventional pruning methods often treat data heterogeneity as a problem to be mitigated. In this work, we introduce a paradigm shift: we reframe client diversity as a feature to be harnessed. We propose AutoFLIP, a framework that begins not with training, but with a one-time federated loss exploration. During this phase, clients collaboratively build a map of the collective loss landscape, using their diverse data to reveal the problem’s essential structure. This shared intelligence then guides an adaptive pruning strategy that is dynamically refined by client agreement throughout training. This approach allows AutoFLIP to identify robust and efficient sub-networks from the outset. Our extensive experiments show that AutoFLIP reduces computational overhead by an average of 52% and communication costs by over 65% while simultaneously achieving state-of-the-art accuracy in challenging non-IID settings.

  • Secure and Privacy-preserving Federated Learning with Explainable Artificial Intelligence for Smart Healthcare System

    by Ali Raza

    Apprentissage fédéré sécurisé et préservant la confidentialité avec intelligence artificielle explicable pour les systèmes intelligents dans le domaine de la santé: La croissance de la population à travers le monde a un impact significatif sur divers secteurs, notamment la main-d'œuvre, les soins de santé et l'économie mondiale. Le secteur de la santé est l'un des secteurs les plus touchés par la croissance de la population en raison de la demande croissante en ressources et en installations de santé. Des systèmes intelligents ont été intégrés pour améliorer la prise de décision afin de faire face à ces problèmes et d'offrir des soins de santé améliorés. Parmi ces systèmes, ceux basés sur l'apprentissage en profondeur (DL) ont surpassé de nombreux systèmes statistiques et ML traditionnels grâce à leur capacité à découvrir et à apprendre automatiquement des caractéristiques liées à une tâche donnée. Par conséquent, l'utilisation du DL a connu une augmentation constante dans de nombreuses applications. Néanmoins, généralement, la formation des modèles DL repose sur un seul serveur centralisé, ce qui pose de nombreux défis : (1) à l'exception de quelques grandes entreprises, la plupart des petites entreprises disposent de données de qualité limitée, insuffisantes pour soutenir la formation des modèles DL, (2) l'accès aux données soulève souvent des problèmes de confidentialité. La collecte et l'analyse d'informations sensibles sur les patients doivent être sécurisées et respectueuses de la vie privée pour protéger les droits individuels à la vie privée, (3) coût élevé de communication et de ressources de calcul requis, (4) un grand nombre de paramètres entraînables rendent les résultats du DL difficiles à expliquer, ce qui est requis dans certaines applications.Par rapport au DL centralisé, l'apprentissage fédéré (FL) améliore à la fois la confidentialité et les coûts de communication, où les clients collaborent pour former un modèle commun sans partager directement les données brutes. Le FL réduit au minimum les violations de la confidentialité et protège les données sensibles en les maintenant distribuées localement. Cela permet une formation collaborative du modèle tout en réduisant le risque d'accès non autorisé et de violations de données. De plus, il favorise l'apprentissage collaboratif en ne partageant que les mises à jour du modèle plutôt que l'ensemble des données.Cependant, le FL présente ses propres défis. Par exemple, la nature hétérogène des données locales entre les clients rend difficile la formation d'un modèle global performant et robuste. De plus, la nature distribuée et le contrôle d'accès des données locales dans le FL le rendent plus vulnérable aux attaques malveillantes. De plus, le défi d'expliquer les résultats du DL reste toujours un défi, et des méthodes doivent être développées pour apporter confiance, responsabilité et transparence dans des applications sensibles, telles que les soins de santé.Par conséquent, l'objectif de cette thèse est de créer des cadres robustes, sécurisés, performants et respectueux de la vie privée dans des environnements fédérés. Ces cadres seront spécifiquement conçus pour des applications de santé de bout en bout, en tenant compte de la présence de données non identiquement distribuées entre


  • Open Access

    Proof of Swarm Based Ensemble Learning for Federated Learning Applications

    by Ali Raza, Kim Phuc Tran, Ludovic Koehl, Shujun Li

    Ensemble learning combines results from multiple machine learning models in order to provide a better and optimised predictive model with reduced bias, variance and improved predictions. However, in federated learning it is not feasible to apply centralised ensemble learning directly due to privacy concerns. Hence, a mechanism is required to combine results of local models to produce a global model. Most distributed consensus algorithms, such as Byzantine fault tolerance (BFT), do not normally perform well in such applications. This is because, in such methods predictions of some of the peers are disregarded, so a majority of peers can win without even considering other peers' decisions. Additionally, the confidence score of the result of each peer is not normally taken into account, although it is an important feature to consider for ensemble learning. Moreover, the problem of a tie event is often left un-addressed by methods such as BFT. To fill these research gaps, we propose PoSw (Proof of Swarm), a novel distributed consensus algorithm for ensemble learning in a federated setting, which was inspired by particle swarm based algorithms for solving optimisation problems. The proposed algorithm is theoretically proved to always converge in a relatively small number of steps and has mechanisms to resolve tie events while trying to achieve sub-optimum solutions. We experimentally validated the performance of the proposed algorithm using ECG classification as an example application in healthcare, showing that the ensemble learning model outperformed all local models and even the FL-based global model.


  • CAPTCHA-Based Secret-Key Sharing Using Quantum Communication

    by KyungHyun Han, Ali Raza, Seong Oun Hwang

    Digital communication has revolutionized the modern era, where two parties can communicate apart easily; but this also opens up opportunities for adversarial attacks. To guarantee a secure communication over such channels, there have been several approaches proposed based on secret-key distribution including the BB84 quantum secret-key sharing protocol. Although it is almost resistant to eavesdropping attacks, the BB84 protocol presents certain challenges. Specifically, it is vulnerable to impersonation attacks. Moreover, it is slow owing to the hit-and-trail method for secret-key distribution. Hence, in this article, we propose a secure and efficient protocol for secret-key sharing by combining image-based CAPTCHA and the classical BB84 protocol. We demonstrate that the proposed protocol is secure against impersonation attacks, including DDoS attacks, and it is at least twice as efficient as BB84 in terms of communicated qubits.


  • Open Access

    Blind Electronic Mail System

    by Maham Imtiaz, Samina Khalid, Saleema Khadam, Sumaira Arshad, Ali Raza, Tehmina Khalil

    Web has gotten one of the essential civilities for everyday living. Each individual is generally getting to the information specially E-mail and data through Web. Nonetheless, daze individuals face challenges in getting to these content materials, likewise in utilizing any help given through Web. The progression in PC based available frameworks has opened up numerous roads for the outwardly debilitated over the globe in a wide manner. Sound input based virtual climate like, the screen readers have helped Blind individuals to get to Web applications colossally. We depict the Voice-mail framework design that can be utilized by a Blind individual to get to messages effectively and productively. The commitment built by this examination has empowered the visually impaired public to deliver and get voice-based e-mail messages in their local language with the assistance of a PC (Ingle, P., Kanade, H., & Lanke, A. 2016).


  • Open Access

    A Framework for Privacy Preserving, Distributed Search Engine Using Topology of DLT and Onion Routing

    by Ali Raza, Kyunghyun Han, Seong Oun Hwang

    Risks regarding Internet privacy and security have been identified as issues for both new and experienced users of Internet technology. This paper explores the privacy and security threats in centralized search engines and proposes a decentralized search system. The proposed decentralized search system provides privacy and security for its users. The decentralized nature also provides transparency, distributed search and community-driven decisions in the proposed framework. We compare the proposed framework with the existing centralized approach. Our analysis shows that (1) there is no significant difference in search time between the centralized and proposed framework, and (2) the proposed framework is more efficient by providing search results with fewer search resources compared to the centralized approach.