ABAymen Boudguiga

Aymen Boudguiga

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Aymen Boudguiga has a total of 68 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    HEGIDE: A MIMD Oblivious Processor for Private Function Evaluation over CKKS

    by Jules Dumezy, Nicolas Ye, Pierre-Emmanuel Clet, Olive Chakraborty, Aymen Boudguiga

  • Threshold (T)FHE Without Smudging by Means of Correct Threshold Additive HE

    by Antonina Bondarchuk, Renaud Sirdey, Aymen Boudguiga, Olive Chakraborty

  • Authentification in wireless mesh networks with identity-based cryptography

    by Aymen Boudguiga

    Authentification dans les réseaux maillés sans fils avec la cryptographie basée sur l’identité: De nos jours, l'authentification dans les réseaux maillés sans fils fait appel aux certificats ou aux secrets partagés. Dans les environnements sans fils, la gestion des certificats est désavantageuse. En effet, les certificats nécessitent le déploiement d'une infrastructure à clés publiques (ICP) et la définition d'une autorité de certification (AC). La AC définit toute une politique qui permet de contrôler la génération, la transmission et la révocation des certificats. Cette politique ne prend pas en considération les limites en termes de puissance et de mémoire que peuvent avoir les stations des clients dans un réseau maillé. Afin de ne pas utiliser les certificats et ne pas déployer une ICP, nous avons étudié dans cette thèse les utilisations possibles de la cryptographie basée sur l’identité (CBI) pour la définition de nouveaux schémas d’authentification pour les réseaux maillés sans fils. La CBI propose de dériver, directement, la clé publique d’une station à partir de son identité. Par conséquent, nous n’avons plus besoin de passer par des certificats pour associer l’identité de la station à sa paire de clés (publique et privée). Par contre, la CBI définit un générateur de clé privée (GCP) qui gère le calcul des clés privées des différentes stations sur le réseau. Par conséquent, ce GCP est capable de réaliser une attaque d’usurpation d’identité (escroc de clés) à l’encontre de toutes les stations légitimes. Pour diminuer le risque de cette attaque, les chercheurs ont tendance à supposer que le GCP est digne de confiance. Dans cette thèse, nous présentons tout d'abord un protocole d'authentification basée sur l’utilisation conjointe d’un mot de passe et de la CBI. En effet, nous proposons d'utiliser le serveur d’authentification de notre réseau maillé comme GCP. Ensuite, nous étudions une liste de mécanismes qui permettent de contrer l’attaque de l’escroc qui caractérise le GCP

  • Open Access

    Robust Federated Learning via Byzantine Filtering over Encrypted Updates

    by Adda Akram Bendoukha, Aymen Boudguiga, Nesrine Kaaniche, Renaud Sirdey, Didem Demirag, Sébastien Gambs

    Federated Learning (FL) aims to train a collaborative model while preserving data privacy. However, the distributed nature of this approach still raises privacy and security issues, such as the exposure of sensitive data due to inference attacks and the influence of Byzantine behaviors on the trained model. In particular, achieving both secure aggregation and Byzantine resilience remains challenging, as existing solutions often address these aspects independently. In this work, we propose to address these challenges through a novel approach that combines homomorphic encryption for privacy-preserving aggregation with property-inference-inspired meta-classifiers for Byzantine filtering. First, following the property-inference attacks blueprint, we train a set of filtering meta-classifiers on labeled shadow updates, reproducing a diverse ensemble of Byzantine misbehaviors in FL, including backdoor, gradient-inversion, label-flipping and shuffling attacks. The outputs of these meta-classifiers are then used to cancel the Byzantine encrypted updates by reweighting. Second, we propose an automated method for selecting the optimal kernel and the dimensionality hyperparameters with respect to homomorphic inference, aggregation constraints and efficiency over the CKKS cryptosystem. Finally, we demonstrate through extensive experiments the effectiveness of our approach against Byzantine participants on the FEMNIST, CIFAR10, GTSRB, and acsincome benchmarks. More precisely, our SVM filtering achieves accuracies between $90$% and $94$% for identifying Byzantine updates at the cost of marginal losses in model utility and encrypted inference runtimes ranging from $6$ to $24$ seconds and from $9$ to $26$ seconds for an overall aggregation.

  • Open Access

    Evaluating Larger Lookup Tables using CKKS

    by Jules Dumezy, Andreea Alexandru, Yuriy Polyakov, Pierre-Emmanuel Clet, Olive Chakraborty, Aymen Boudguiga

    The Cheon–Kim–Kim–Song (CKKS) scheme is a fully homomorphic encryption scheme that traditionally supports only the evaluation of smooth functions. Recent works have enabled the evaluation of arbitrary (discontinuous) integer functions represented as lookup tables (LUT) on small inputs using the method of functional bootstrapping (FBT). Although well-suited for small integers (up to around 10 bits), the efficiency of FBT quickly declines for large LUTs, and a considerable increase in both runtime and memory requirements is observed. Building on CKKS functional bootstrapping, we propose in this paper two functional bootstrapping algorithms, specifically designed to target larger LUTs (up to 20 bits). For a 16-bit LUT, our implementation in OpenFHE achieves a speed-up of 47.5 in amortized time and 95.1 in latency for single-threaded execution, compared to the state-of-the-art CKKS-based functional bootstrapping method of Alexandru et al. (CRYPTO’25).


  • Open Access

    Running Standard Block Ciphers Beyond AES with TFHE: Experiments and Lessons Learnt

    by Daphné Trama, Aymen Boudguiga, Renaud Sirdey

    The dream of achieving data privacy during external computations has become increasingly concrete in recent years. Indeed, since the early days of Fully Homomorphic Encryption (FHE) more than a decade ago, new cryptosystems and techniques have constantly optimized the efficiency of computation on encrypted data. However, one of the main disadvantages of FHE, namely its significant ciphertext expansion factor, remains at the center of the efficiency bottleneck of FHE schemes. To tackle the issue of slow uplink FHE data transmission, we use transciphering. With transciphering, the client naturally encrypts its data under a symmetric scheme and sends them to the server with (once and for all) an FHE encryption of the symmetric scheme's key. With its larger computing power, the server then evaluates the symmetric scheme's decryption algorithm within the homomorphic domain to obtain homomorphic ciphertexts that allow it to perform the requested calculations. Since the first use of this method a bit more than ten years ago, papers on the homomorphic evaluation of AES have been numerous. And as the AES execution is the application chosen by NIST in the FHE part of its recent call for proposals on threshold encryption, the stakes of such work go up another level. But what about other standardized block ciphers? Is the AES the more efficient option? In this work, we leverage on two methods which have successfully been applied to the homomorphic evaluation of AES to study several state-of-the-art symmetric block ciphers (namely CLEFIA, PRESENT, PRINCE, SIMON, SKINNY). That is to say, we implement a representative set of symmetric block ciphers using TFHE. These implementations allow us to compare the efficiency of this set of symmetric schemes and to categorize them. We highlight the characteristics of block ciphers that are fast to execute in the homomorphic domain and those that are particularly costly. Finally, this classification of operation types enables us to sketch out what the ideal block cipher for transciphering homomorphic data in integer mode might look like.


  • Open Access

    SplitML: A Unified Privacy-Preserving Architecture for Federated Split-Learning in Heterogeneous Environments

    by Devharsh Trivedi, Aymen Boudguiga, Nesrine Kaaniche, Nikos Triandopoulos

    While Federated Learning (FL) and Split Learning (SL) aim to uphold data confidentiality by localized training, they remain susceptible to adversarial threats such as model poisoning and sophisticated inference attacks. To mitigate these vulnerabilities, we propose SplitML, a secure and privacy-preserving framework for Federated Split Learning (FSL). By integrating IND−CPAD secure Fully Homomorphic Encryption (FHE) with Differential Privacy (DP), SplitML establishes a defense-in-depth strategy that minimizes information leakage and thwarts reconstructive inference attempts. The framework accommodates heterogeneous model architectures by allowing clients to collaboratively train only the common top layers while keeping their bottom layers exclusive to each participant. This partitioning strategy ensures that the layers closest to the sensitive input data are never exposed to the centralized server. During the training phase, participants utilize multi-key CKKS FHE to facilitate secure weight aggregation, which ensures that no single entity can access individual updates in plaintext. For collaborative inference, clients exchange activations protected by single-key CKKS FHE to achieve a consensus derived from Total Labels (TL) or Total Predictions (TP). This consensus mechanism enhances decision reliability by aggregating decentralized insights while obfuscating soft-label confidence scores that could be exploited by attackers. Our empirical evaluation demonstrates that SplitML provides substantial defense against Membership Inference (MI) attacks, reduces temporal training costs compared to standard encrypted FL, and improves inference precision via its consensus mechanism, all while maintaining a negligible impact on federation overhead.


  • Open Access

    SplitML: A Unified Privacy-Preserving Architecture for Federated Split-Learning in Heterogeneous Environments

    by Devharsh Trivedi, Aymen Boudguiga, Nesrine Kaaniche, Nikos Triandopoulos

    Federated Learning (FL) and Split Learning (SL) maintain client data privacy during collaborative training by keeping raw data on distributed clients and only sharing model updates (FL) or intermediate results (SL) with the centralized server. However, this level of privacy is insufficient, as both FL and SL remain vulnerable to security risks like poisoning and various inference attacks. To address these flaws, we introduce SplitML, a secure and privacy-preserving framework for Federated Split Learning (FSL). SplitML generalizes and formalizes FSL using IND−CPAD secure Fully Homomorphic Encryption (FHE) combined with Differential Privacy (DP) to actively reduce data leakage and inference attacks. This framework allows clients to use different overall model architectures, collaboratively training only the top (common) layers while keeping their bottom layers private. For training, clients use multi-key CKKS FHE to aggregate weights. For collaborative inference, clients can share gradients encrypted with single-key CKKS FHE to reach a consensus based on Total Labels (TL) or Total Predictions (TP). Empirical results show that SplitML significantly improves protection against Membership Inference (MI) attacks, reduces training time, enhances inference accuracy through consensus, and incurs minimal federation overhead.


  • Open Access

    FairCognizer: A Model for Accurate Predictions with Inherent Fairness Evaluation (Extended Abstract)

    by Adda Akram Bendoukha, Nesrine Kaaniche, Aymen Boudguiga, Renaud Sirdey

    Algorithmic fairness is a critical challenge in building trustworthy Machine Learning (ML) models. ML classifiers strive to make predictions that closely match real-world observations (ground truth). However, if the ground truth data itself reflects biases against certain sub-populations, a dilemma arises: prioritize fairness and potentially reduce accuracy, or emphasize accuracy at the expense of fairness. This work proposes a novel training framework that goes beyond achieving high accuracy. Our framework trains a classifier to not only deliver optimal predictions but also to identify potential fairness risks associated with each prediction. To do so, we specify a dual-labeling strategy where the second label contains a per-prediction fairness evaluation, referred to as an unfairness risk evaluation. In addition, we identify a subset of samples as highly vulnerable to group-unfair classifiers. Our experiments demonstrate that our classifiers attain optimal accuracy levels on both the Adult-Census-Income and Compas-Recidivism datasets. Moreover, they identify unfair predictions with nearly 75% accuracy at the cost of expanding the size of the classifier by 45%.

  • Open Access

    FADE: Federated Aggregation with Discrimination Elimination

    by Adda-Akram Bendoukha, Héber Hwang Arcolezi, Nesrine Kaaniche, Aymen Boudguiga, Renaud Sirdey, Pierre-Emmanuel Clet