Dissertations / Theses on the topic 'Chiffrement homomorphe (informatique)'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 30 dissertations / theses for your research on the topic 'Chiffrement homomorphe (informatique).'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Tap, Samuel. "Construction de nouveaux outils de chiffrement homomorphe efficace." Electronic Thesis or Diss., Université de Rennes (2023-....), 2023. http://www.theses.fr/2023URENS103.
Full textIn our everyday life, we leave a trail of data whenever we access online services. Some are given voluntarily and others reluctantly. Those data are collected and analyzed in the clear which leads to major threats on the user's privacy and prevents collaborations between entities working on sensitive data. In this context, Fully Homomorphic Encryption brings a new hope by enabling computation over encrypted data, which removes the need to access data in the clear to analyze and exploit it. This thesis focuses on TFHE, a recent fully homomorphic encryption scheme able to compute a bootstrapping in record time. We introduce an optimization framework to set the degrees of freedom inherent to homomorphic computations which gives non-experts the ability to use it (more) easily. We describe a plethora of new FHE algorithms which improve significantly the state of the art and limit, (if not remove) existing restrictions. Efficient open source implementations are already accessible
Barrier, Joris. "Chiffrement homomorphe appliqué au retrait d'information privé." Thesis, Toulouse, INSA, 2016. http://www.theses.fr/2016ISAT0041/document.
Full textPrivate information retrieval, named PIR, is a set of protocols that is a part of privacy enhancement technologies.Its major feature is to hide the index of a record that a user retrieved from the host.Without neglecting the scientific contributions of its authors, the usability of this protocol seems hard since that, for a user, it seems more and more efficient to receive all the records.Thus far, PIR can be achieved using mutually distrustful databases replicated databases, trusted hardware, or cryptographic systems.We focus on computational private information retrieval, and specifically on thus based on cryptographic systems.This decision is contingent to the spread of cryptographic systems based on lattices who provide specific properties.To demonstrate it usability, we offer an efficient and easy-to-use private Information retrieval based on homomorphic encryption
Paindavoine, Marie. "Méthodes de calculs sur les données chiffrées." Thesis, Lyon, 2017. http://www.theses.fr/2017LYSE1009/document.
Full textNowadays, encryption and services issued of ``big data" are at odds. Indeed, encryption is about protecting users privacy, while big data is about analyzing users data. Being increasingly concerned about security, users tend to encrypt their sensitive data that are subject to be accessed by other parties, including service providers. This hinders the execution of services requiring some kind of computation on users data, which makes users under obligation to choose between these services or their private life. We address this challenge in this thesis by following two directions.In the first part of this thesis, we study fully homomorphic encryption that makes possible to perform arbitrary computation on encrypted data. However, this kind of encryption is still inefficient, and this is due in part to the frequent execution of a costly procedure throughout evaluation, namely the bootstrapping. Thus, efficiency is inversely proportional to the number of bootstrappings needed to evaluate functions on encrypted data. In this thesis, we prove that finding such a minimum is NP-complete. In addition, we design a new method that efficiently finds a good approximation of it. In the second part, we design schemes that allow a precise functionality. The first one is verifiable deduplication on encrypted data, which allows a server to be sure that it keeps only one copy of each file uploaded, even if the files are encrypted, resulting in an optimization of the storage resources. The second one is intrusion detection over encrypted traffic. Current encryption techniques blinds intrusion detection services, putting the final user at risks. Our results permit to reconcile users' right to privacy and their need of keeping their network clear of all intrusion
Migliore, Vincent. "Cybersécurite matérielle et conception de composants dédiés au calcul homomorphe." Thesis, Lorient, 2017. http://www.theses.fr/2017LORIS456/document.
Full textThe emergence of internet and the improvement of communica- tion infrastructures have considerably increased the information flow around the world. This development has come with the emergence of new needs and new expectations from consumers. Communicate with family or colleagues, store documents or multimedia files, using innovative services which processes our personal data, all of this im- plies sharing with third parties some potentially sensitive data. If third parties are untrusted, they can manipulate without our agreement data we share with them. In this context, homomorphic encryption can be a good solution. Ho- momorphic encryption can hide to the third parties the data they are processing. However, at this point, homomorphic encryption is still complex. To process a few bits of clear data (cleartext), one needs to manage a few million bits of encrypted data (ciphertext). Thus, a computation which is usually simple becomes very costly in terms of computation time. In this work, we have improved the practicability of homomorphic en- cryption by implementing a specific accelerator. We have followed a software/hardware co-design approach with the help of Karatsuba algorithm. In particular, our approach is compatible with batching, a technique that “packs" several messages into one ciphertext. Our work demonstrates that the batching can be implemented at no important additional cost compared to non-batching approaches, and allows both reducing computation time (operations are processed in parallel) and the ciphertext/cleartext ratio
Chen, Yuanmi. "Réduction de réseau et sécurité concrète du chiffrement complètement homomorphe." Paris 7, 2013. http://www.theses.fr/2013PA077242.
Full textThe popularity of lattice-based cryptography has significantly increased in the past few years with the discovery of new spectacular functionalities such as fully-homomorphic encryption and (indistinguishability) obfuscation. It has become crucial to be able to analyze the concrete security of lattice-based cryptosystems, in order to select their parameters and to assess their practical performances. In a first part, we present a theoretical analysis and concrete improvements to the so-called BKZ reduction, which is considered tô be the most efficient lattice reduction algorithm in practice for high dimensions. We begin by studying the main subroutine of BKZ, enumeration, and we extend the analysis of pruned enumeration by Gama, Nguyen and Regev (EUROCRYPT 2010). Next, we improve the BKZ algorithm by using several techniques, such as pruned enumeration, pre-processing and abort. And we discuss how to select BKZ parameters efficiently. Based on numerous experiments, we present a simulation algorithm to predict the output quality of BKZ reduction. This allows us to revise the security estimates of numerous lattice-based cryptosystems, and explain how to solve SVP by enumeration as efficiently as possible, based on the state-of-the-art. In a second part, we present a new algorithm for the approximate greatest common divisor problem, using a time/memory trade-off. This provides a better concrete attack on the fully-homomorphic encryption scheme proposed by Coron, Mandal, Naccache and Tibouchi (CRYPTO 2011). It also has other applications in cryptanalysis
Chinthamani, Dwarakanath Nagarjun. "Theoretical and practical contributions to homomorphic encryption." Electronic Thesis or Diss., université Paris-Saclay, 2021. http://www.theses.fr/2021UPASG103.
Full textIn conventional encryption schemes, the primary aim of the scheme is to ensure confidentiality of the data. Fully Homomorphic Encryption (FHE), a variant first realized by Gentry, is an encryption scheme which also allows for computation over the encrypted data, without ever needing to decrypt it. Using this, any untrusted third party with the relevant key material can perform homomorphic computations, leading to many applications where an untrusted party can still be allowed to compute over encryptions of sensitive data (cloud computing), or where the trust needs to be decentralized (multi-party computation).This thesis consists of two main contributions to Fully Homomorphic Encryption. In the first part, we take an FHE based on Fermat numbers and extend it to work with multi-bit numbers. We also add the ability to homomorphically evaluate small functions, with which we can compute additions and multiplication with only a few bootstrappings, and these can be used as building blocks for larger computations. Some newer results on sub-Gaussian random variables are adapted to give a better error analysis.One of the obstacles in bringing FHE to the mainstream remains its large computational complexity, and optimized architectures to accelerate FHE computations on reconfigurable hardware have been proposed. The second part of our thesis proposes an architecture for the polynomial arithmetic used in FV-like cryptosystems. This can be used to compute the sum and product of ring polynomials, using a pair of NTT algorithms which avoids the use of bit reversal, and subsumes the need for multiplication by weight vectors. For the cost of storing twiddle factors in a ROM, we avoid twiddle updates leading to a much smaller cycle count
Méaux, Pierrick. "Hybrid fully homomorphic framework." Thesis, Paris Sciences et Lettres (ComUE), 2017. http://www.theses.fr/2017PSLEE066/document.
Full textFully homomorphic encryption, firstly built in 2009, is a very powerful kind of encryption, allowing to compute any function on encrypted data, and to get an encrypted version of the result. Such encryption enables to securely delegate data to a cloud, ask for computations, recover the result, while keeping private the data during the whole process. However, today’s inefficiency of fully homomorphic encryption, and its inadequateness to the outsourcing computation context, makes its use alone insufficient for this application. Both of these issues can be circumvented, using fully homomorphic encryption in a larger framework, by combining it with a symmetric encryption scheme. This combination gives a hybrid fully homomorphic framework, designed towards efficient outsourcing computation, providing both security and privacy. In this thesis, we contribute to the study of hybridfully homomorphic framework, through the analysis, and the design of symmetric primitives making efficient this hybrid construction. Through the examination of fully homomorphic encryption schemes, we develop tools to efficiently use the homomorphic properties in a more complex framework. By investigating various symmetric encryption schemes, and buildingblocks up to the circuit level, we determine good candidates for a hybrid context. Through evaluating the security of constructions optimizing the homomorphic evaluation, we contribute to a wide study within the cryptographic Boolean functions area. More particularly, we introduce a new family of symmetric encryption schemes, with a new design, adapted to the hybrid fully homomorphic framework. We then investigate its behavior relatively to homomorphic evaluation, and we address the security of such design. Finally, particularities of this family of ciphers motivate specific cryptanalyses, therefore we develop and analyze new cryptographic Boolean criteria
Chillotti, Ilaria. "Vers l'efficacité et la sécurité du chiffrement homomorphe et du cloud computing." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLV020.
Full textFully homomorphic encryption is a new branch of cryptology, allowing to perform computations on encrypted data, without having to decrypt them. The main interest of homomorphic encryption schemes is the large number of practical applications for which they can be used. Examples are given by electronic voting, computations on sensitive data, such as medical or financial data, cloud computing, etc..The first fully homomorphic encryption scheme has been proposed in 2009 by Gentry. He introduced a new technique, called bootstrapping, used to reduce the noise in ciphertexts: in fact, in all the proposed homomorphic encryption schemes, the ciphertexts contain a small amount of noise, which is necessary for security reasons. If we perform computations on noisy ciphertexts, the noise increases and, after a certain number of operations, the noise becomes to large and it could compromise the correctness of the final result, if not controlled.Bootstrapping is then fundamental to construct fully homomorphic encryption schemes, but it is very costly in terms of both memory and time consuming.After Gentry’s breakthrough, the presented schemes had the goal to propose new constructions and to improve bootstrapping, in order to make homomorphic encryption practical. One of the most known schemes is GSW, proposed by Gentry, Sahai et Waters in 2013. The security of GSW is based on the LWE (learning with errors) problem, which is considered hard in practice. The most rapid bootstrapping on a GSW-based scheme has been presented by Ducas and Micciancio in 2015. In this thesis, we propose a new variant of the scheme proposed by Ducas and Micciancio, that we call TFHE.The TFHE scheme improves previous results, by performing a faster bootstrapping (in the range of a few milliseconds) and by using smaller bootstrapping keys, for the same security level. TFHE uses TLWE and TGSW ciphertexts (both scalar and ring): the acceleration of bootstrapping is mainly due to the replacement of the internal GSW product, used in the majority of previous constructions, with an external product between TLWE and TGSW.Two kinds of bootstrapping are presented. The first one, called gate bootstrapping, is performed after the evaluation of a homomorphic gate (binary or Mux); the second one, called circuit bootstrapping, can be executed after the evaluation of a larger number of homomorphic operations, in order to refresh the result or to make it compatible with the following computations.In this thesis, we also propose new techniques to improve homomorphic computations without bootstrapping and new packing techniques. In particular, we present a vertical packing, that can be used to efficiently evaluate look-up tables, we propose an evaluation via weighted deterministic automata, and we present a homomorphic counter, called TBSR, that can be used to evaluate arithmetic functions.During the thesis, the TFHE scheme has been implemented and it is available in open source.The thesis contains also ancillary works. The first one concerns the study of the first model of post-quantum electronic voting based on fully homomorphic encryption, the second one analyzes the security of homomorphic encryption in a practical cloud implementation scenario, and the third one opens up about a different solution for secure computing, multi-party computation
Bonnoron, Guillaume. "A journey towards practical fully homomorphic encryption." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2018. http://www.theses.fr/2018IMTA0073/document.
Full textCraig Gentry presented in 2009 the first fully homomorphic encryption scheme. Since then, a tremendous effort has been, and still is, dedicated by the cryptographic community to make practical this new kind of cryptography. It is revolutionnary because it enables direct computation on encrypted data (without the need for the computing entity to decrypt them). Several trends have been developed in parallel, exploring on one side fully homomorphic encryption schemes, more versatile for applications but more costly in terms of time and memory. On the other side, the somewhat homomorphic encryption schemes are less flexible but more efficient. This thesis, achieved within the Chair of Naval Cyber Defence, contributes to these trends. We have endorsed different roles. First, an attacker position to assess the hardness of the security assumptions of the proposals. Then, we conducted a state-of-the-art of the most promising schemes in order to identify the best(s) depending on the use-cases and to give precise advice to appropriately set the parameters that drive security level, ciphertext sizes and computation costs. Last, we endorsed a designer role. We proposed a new powerful fully homomorphic encryption scheme together with its open-source implementation, available on github
Belhadj, Djedjiga. "Multi-GAT semi-supervisé pour l’extraction d’informations et son adaptation au chiffrement homomorphe." Electronic Thesis or Diss., Université de Lorraine, 2024. http://www.theses.fr/2024LORR0023.
Full textThis thesis is being carried out as part of the BPI DeepTech project, in collaboration with the company Fair&Smart, primarily looking after the protection of personal data in accordance with the General Data Protection Regulation (RGPD). In this context, we have proposed a deep neural model for extracting information in semi-structured administrative documents (SSDs). Due to the lack of public training datasets, we have proposed an artificial generator of SSDs that can generate several classes of documents with a wide variation in content and layout. Documents are generated using random variables to manage content and layout, while respecting constraints aimed at ensuring their similarity to real documents. Metrics were introduced to evaluate the content and layout diversity of the generated SSDs. The results of the evaluation have shown that the generated datasets for three SSD types (payslips, receipts and invoices) present a high diversity level, thus avoiding overfitting when training the information extraction systems. Based on the specific format of SSDs, consisting specifically of word pairs (keywords-information) located in spatially close neighborhoods, the document is modeled as a graph where nodes represent words and edges, neighborhood connections. The graph is fed into a multi-layer graph attention network (Multi-GAT). The latter applies the multi-head attention mechanism to learn the importance of each word's neighbors in order to better classify it. A first version of this model was used in supervised mode and obtained an F1 score of 96% on two generated invoice and payslip datasets, and 89% on a real receipt dataset (SROIE). We then enriched the multi-GAT with multimodal embedding of word-level information (textual, visual and positional), and combined it with a variational graph auto-encoder (VGAE). This model operates in semi-supervised mode, being able to learn on both labeled and unlabeled data simultaneously. To further optimize the graph node classification, we have proposed a semi-VGAE whose encoder shares its first layers with the multi-GAT classifier. This is also reinforced by the proposal of a VGAE loss function managed by the classification loss. Using a small unlabeled dataset, we were able to improve the F1 score obtained on a generated invoice dataset by over 3%. Intended to operate in a protected environment, we have adapted the architecture of the model to suit its homomorphic encryption. We studied a method of dimensionality reduction of the Multi-GAT model. We then proposed a polynomial approximation approach for the non-linear functions in the model. To reduce the dimensionality of the model, we proposed a multimodal feature fusion method that requires few additional parameters and reduces the dimensions of the model while improving its performance. For the encryption adaptation, we studied low-degree polynomial approximations of nonlinear functions, using knowledge distillation and fine-tuning techniques to better adapt the model to the new approximations. We were able to minimize the approximation loss by around 3% on two invoice datasets as well as one payslip dataset and by 5% on SROIE
Madi, Abbass. "Secure Machine Learning by means of Homomorphic Encryption and Verifiable Computing." Electronic Thesis or Diss., université Paris-Saclay, 2022. http://www.theses.fr/2022UPASG019.
Full textMachine Learning (ML) represents a new trend in science because of its power to solve problems automatically and its wide spectrum of applications (e.g., business, healthcare domain, etc.). This attractive technology caught our attention from a cryptography point of view in the sense that we worked during this Ph.D. to ensure secure usage of ML setups. Our Ph.D. work proposes a secure remote evaluation over different ML setups (for inference and for training). This thesis addresses two cornerstones: confidentiality of training or inference data and remote evaluation integrity in different ML setups (federated learning or transfer learning-based inference). In contrast, most other works focus only on data confidentiality. In our thesis, we proposed three architectures/frameworks to ensure a secure remote evaluation for the following ML setups: Neural Networks (NN), Federated Learning (FL), and Transfer Learning (TL). Particularly, our FL and TL architectures are the first that treat both the confidentiality and integrity security properties. We built these architectures using or modifying pre-existing VC schemes over homomorphic encrypted data: mainly we use VC protocols for BFV and Paillier schemes. An essential characteristic for our architectures is their generality, in the sense, if there are improvements in VC protocols and HE schemes, our frameworks can easily take into account these new approaches. This work opens up many perspectives, not only in privacy-preserving ML architectures, but also for the tools used to ensure the security properties. For example, one important perspective is to add differential privacy (DP) to our FL architecture
Zucca, Vincent. "Towards efficient arithmetic for Ring-LWE based homomorphic encryption." Electronic Thesis or Diss., Sorbonne université, 2018. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2018SORUS080.pdf.
Full textFully homomorphic encryption is a kind of encryption offering the ability to manipulate encrypted data directly through their ciphertexts. In this way it is possible to process sensitive data without having to decrypt them beforehand, ensuring therefore the datas' confidentiality. At the numeric and cloud computing era this kind of encryption has the potential to considerably enhance privacy protection. However, because of its recent discovery by Gentry in 2009, we do not have enough hindsight about it yet. Therefore several uncertainties remain, in particular concerning its security and efficiency in practice, and should be clarified before an eventual widespread use. This thesis deals with this issue and focus on performance enhancement of this kind of encryption in practice. In this perspective we have been interested in the optimization of the arithmetic used by these schemes, either the arithmetic underlying the Ring Learning With Errors problem on which the security of these schemes is based on, or the arithmetic specific to the computations required by the procedures of some of these schemes. We have also considered the optimization of the computations required by some specific applications of homomorphic encryption, and in particular for the classification of private data, and we propose methods and innovative technics in order to perform these computations efficiently. We illustrate the efficiency of our different methods through different software implementations and comparisons to the related art
Ameur, Yulliwas. "Exploring the Scope of Machine Learning using Homomorphic Encryption in IoT/Cloud." Electronic Thesis or Diss., Paris, HESAM, 2023. http://www.theses.fr/2023HESAC036.
Full textMachine Learning as a Service (MLaaS) has accelerated the adoption of machine learning techniques in various domains. However, this trend has also raised serious concerns over the security and privacy of the sensitive data used in machine learning models. To address this challenge, our approach is to use homomorphic encryption.The aim of this thesis is to examine the implementation of homomorphic encryption in different applications of machine learning.. The first part of the work focuses on the use of homomorphic encryption in a multi-cloud environment, where the encryption is applied to simple operations such as addition and multiplication.This thesis explores the application of homomorphic encryption to the k-nearest neighbors (k-NN) algorithm. The study presents a practical implementation of the k-NN algorithm using homomorphic encryption and demonstrates the feasibility of this approach on a variety of datasets. The results show that the performance of the k-NN algorithm using homomorphic encryption is comparable to that of the unencrypted algorithm.Third, the work investigates the application of homomorphic encryption to the k-means clustering algorithm. Similar to the k-NN study, the thesis presents a practical implementation of the k-means algorithm using homomorphic encryption and evaluates its performance on various datasets.Finally, the thesis explores the combination of homomorphic encryption with differential privacy (DP) techniques to further enhance the privacy of machine learning models. The study proposes a novel approach that combines homomorphic encryption with DP to achieve better privacy guarantees for machine learning models. The research presented in this thesis contributes to the growing body of research on the intersection of homomorphic encryption and machine learning, providing practical implementations and evaluations of homomorphic encryption in various machine learning contexts.iffalseAccording to Gartner, 5.8 Billion Enterprise and Automotive IoT endpoints will be in use at the end of 2020 while Statistica shows that IoT enablers solutions (such as Cloud, analytics, security) will reach 15 Billion of euros in the European Union market by 2025. However, these IoT devices have not enough resource capacity to process the data collected by their sensors making these devices vulnerable and prone to attack. To avoid processing data within the IoT devices, the trend is to outsource the sensed data to the Cloud that has both resourceful data storage and data processing. Nevertheless, the externalized data may be sensitive, and the users may lose privacy on the data content while allowing the cloud providers to access and possibly use these data to their own business. To avoid this situation and preserve data privacy in the Cloud datacenter, one possible solution is to use the fully homomorphic encryption (FHE) that assures both confidentiality and efficiency of the processing. In many smart environments such as smart cities, smart health, smart farming, industry 4.0, etc. where massive data are generated, there is a need to apply machine learning (ML) techniques, hence contributing to the decision making to act on the smart environment. Indeed, the challenging issue in this context is to adapt the ML approaches to apply them on encrypted data so that the decision taken on encrypted data can be reported on the cleartext data. This PhD thesis is a cooperative research work between two teams ROC and MSDMA of CEDRIC Lab. It aims at exploring the use of ML and FHE in smart applications where IoT devices collect sensitive data to outsource them on untrusted Cloud datacenter for computing thanks to ML models
Grivet, Sébert Arnaud. "Combining differential privacy and homomorphic encryption for privacy-preserving collaborative machine learning." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASG037.
Full textThe purpose of this PhD is to design protocols to collaboratively train machine learning models while keeping the training data private. To do so, we focused on two privacy tools, namely differential privacy and homomorphic encryption. While differential privacy enables to deliver a functional model immune to attacks on the training data privacy by end-users, homomorphic encryption allows to make use of a server as a totally blind intermediary between the data owners, that provides computational resource without any access to clear information. Yet, these two techniques are of totally different natures and both entail their own constraints that may interfere: differential privacy generally requires the use of continuous and unbounded noise whereas homomorphic encryption can only deal with numbers encoded with a quite limited number of bits. The presented contributions make these two privacy tools work together by coping with their interferences and even leveraging them so that the two techniques may benefit from each other.In our first work, SPEED, we built on Private Aggregation of Teacher Ensembles (PATE) framework and extend the threat model to deal with an honest but curious server by covering the server computations with a homomorphic layer. We carefully define which operations are realised homomorphically to make as less computation as possible in the costly encrypted domain while revealing little enough information in clear to be easily protected by differential privacy. This trade-off forced us to realise an argmax operation in the encrypted domain, which, even if reasonable, remained expensive. That is why we propose SHIELD in another contribution, an argmax operator made inaccurate on purpose, both to satisfy differential privacy and lighten the homomorphic computation. The last presented contribution combines differential privacy and homomorphic encryption to secure a federated learning protocol. The main challenge of this combination comes from the necessary quantisation of the noise induced by encryption, that complicates the differential privacy analysis and justifies the design and use of a novel quantisation operator that commutes with the aggregation
Hiscock, Thomas. "Microcontrôleur à flux chiffré d'instructions et de données." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLV074/document.
Full textEmbedded processors are today ubiquitous, dozen of them compose and orchestrate every technology surrounding us, from tablets to smartphones and a large amount of invisible ones. At the core of these systems, processors gather data, process them and interact with the outside world. As such, they are excepted to meet very strict safety and security requirements. From a security perspective, the task is even more difficult considering the user has a physical access to the device, allowing a wide range of specifically tailored attacks.Confidentiality, in terms of both software code and data is one of the fundamental properties expected for such systems. The first contribution of this work is a software encryption method based on the control flow graph of the program. This enables the use of stream ciphers to provide lightweight and efficient encryption, suitable for constrained processors. The second contribution is a data encryption mechanism based on homomorphic encryption. With this scheme, sensible data remain encrypted not only in memory, but also during computations. Then, the integration and evaluation of these solutions on Field Programmable Gate Array (FPGA) with some example programs will be discussed
Meyer, Pierre. "Sublinear-communication secure multiparty computation." Electronic Thesis or Diss., Université Paris Cité, 2023. http://www.theses.fr/2023UNIP7129.
Full textSecure Multi-Party Computation (MPC) [Yao82, GMW87a] allows a set of mutually distrusting parties to perform some joint computation on their private inputs without having to reveal anything beyond the output. A major open question is to understand how strongly the communication complexity of MPC and the computational complexity of the function being computed are correlated. An intriguing starting point is the study of the circuit-size barrier. The relevance of this barrier is a historical, and potentially absolute, one: all seminal protocols from the 1980s and 1990s use a "gate-by-gate" approach, requiring interaction between the parties for each (multiplicative) gate of the circuit to be computed, and this remains the state of the art if we wish to provide the strongest security guarantees. The circuit-size barrier has been broken in the computational setting from specific, structured, computational assumption, via Fully Homomorphic Encryption (FHE) [Gen09] and later Homomorphic Secret Sharing [BGI16a]. Additionally, the circuit-size barrier for online communication has been broken (in the correlated randomness model) information-theoretically [IKM + 13, DNNR17, Cou19], but no such result is known for the total communication complexity (in the plain model). Our methodology is to draw inspiration from known approaches in the correlated randomness model, which we view simultaneously as fundamental (because it provides information-theoretic security guarantees) and inherently limited (because the best we can hope for in this model is to understand the online communication complexity of secure computation), in order to devise new ways to break the circuit-size barrier in the computational setting. In the absence of a better way to decide when concrete progress has been made, we take extending the set of assumptions known to imply sublinear-communication secure computation as "proof of conceptual novelty". This approach has allowed us to break the circuit-size barrier under quasipolynomial LPN [CM21] or QR and LPN [BCM22]. More fundamentally, these works constituted a paradigm shift, away from the "homomorphism-based" approaches of FHE and HSS, which ultimately allowed us to break the two-party barrier for sublinear-communication secure computation and provide in [BCM23] the first sublinear-communication protocol with more than two parties, without FHE. Orthogonally to this line of work, purely focusing on computational security, we showed in [CMPR23] that [BGI16a] could be adapted to provide information-theoretic security for one of the two parties, and computational security for the other: these are provably the strongest security guarantees one can hope to achieve in the two-party setting (without setup), and ours is the first sublinear-communication protocol in this setting which does not use FHE
Zuber, Martin. "Contributions to data confidentiality in machine learning by means of homomorphic encryption." Electronic Thesis or Diss., université Paris-Saclay, 2020. http://www.theses.fr/2020UPASG048.
Full textWe aim to provide a set tools allowing for machine learning algorithms to yield their intended results while ensuring confidentiality properties are achieved for the underlying data. This can be achieved through regulatory measures such as prohibiting the use of a sensitive database in certain cases and restricting its access to certain law enforcement agencies. The fundamental reason for the existence of our work - and every other work like it - is the following: why trust that an outside entity will not misuse personal data when you can have assurances of that fact ? This applies both in the case of a private company that may use/sell your data for profit, legally or illegally. It also applies to use by a government which may or may not have the proper safeguards against abuse, as well as the proper security for data storage and access. In our case, we provide such confidentiality properties though the use of Fully Homomorphic Encryption (FHE). Precisely, most of our work focuses on finding new algorithms for secure outsourced machine learning evaluation using FHE. While other privacy and confidentiality preserving methods are touched upon briefly, we focused our research on homomorphic encryption and strive to explain our choice and its general context. We present three main novel secure machine learning applications: a confidentiality-preserving recursive discrete neural network; a model-confidential embedding-based neural network; a confidentiality-preserving k-NN classifier. Notably, our secure k-NN classifier is the only such algorithm in the literature obtaining a result noninteractively. We evaluate the accuracy and efficiency of these three applications on real-world machine learning problems. We show that our secure schemes compare very favorably to their non-secure counterparts in terms of accuracy, while still running in realistic time. Beyond these schemes themselves, this thesis promotes a specific research direction for secure machine learning. We argue for less (though still some) focus on deep convolutional neural networks and show that looking at somewhat lesser known machine learning algorithms can yield promising results
Fau, Simon. "Systèmes de cryptocalculs, compilation et support d’exécution." Thesis, Lorient, 2016. http://www.theses.fr/2016LORIS398/document.
Full textOur approach in this thesis was to identify where FHE could be used in computer science and to build an experimental platform that allow us to test real-life algorithm running on homomorphically-encrypted data. The first part of this thesis is dedicated to the state of the art. We first present homomorphic encryption schemes designed before 2008 and then move to the Fully Homomorphic Encryption period. We describe several schemes of interest for this thesis and discuss FHE implementations. Finally, we present Yao’s garbled circuits as they can solve similar problems as FHE and briefly talk about Functional Encryption (FE). The second part of this thesis is for our contributions to the subject. We begin by explaining how FHE can be useful in various scenarios and try to provide practical use cases that we identified during the thesis. Then, we describe our approach to perform computations on encrypted data using FHE and explain how we were able to build on just the homomorphic addition and multiplication a platform for the execution in the encrypted domain of a wide range of algorithms. We then detail our solution for performing private queries on an encrypted database using homomorphic encryption. In a final chapter, we present our experimental results
Niyitegeka, David. "Composition de mécanismes cryptographiques et de tatouage pour la protection de données génétiques externalisées." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2020. http://www.theses.fr/2020IMTA0225.
Full textNowadays, cloud computing allows researchers and health professionals to flexibly store and process large amounts of genetic data remotely, without a need to purchase and to maintain their own infrastructures. These data are especially used in genome-wide association studies (GWAS) in order to conduct the identification of genetic variants that are associated with some diseases. However, genetic data outsourcing or sharing in the cloud induces many security issues. In addition, a human genome is very sensitive by nature and represents the unique biological identity of its owner. The objective of this thesis work is to protect genetic data during their sharing, storage and processing. We have developped new security tools that are based on watermarking and cryptographic mechanisms, as well as on the combination of them. First, we have proposed a privacy-preserving method that allows to compute the secure collapsing method based on the logistic regression model using homomorphic encryption (HE). To overcome the computational and storage overhead of HE-based solutions, we have developed a framework that allows secure performing of GWAS for rare variants without increasing complexity compared to its nonsecure version. It is based on several security mechanisms including encryption. In parallel of these works, we have exploited the semantic security of some HE schemes so as to develop a dynamic watermarking method that allows integrity control for encrypted data. At last, we have developed a robust watermarking tool for GWAS data for traitor tracing purposes
Bellafqira, Reda. "Chiffrement homomorphe et recherche par le contenu sécurisé de données externalisées et mutualisées : Application à l'imagerie médicale et l'aide au diagnostic." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2017. http://www.theses.fr/2017IMTA0063.
Full textCloud computing has emerged as a successful paradigm allowing individuals and companies to store and process large amounts of data without a need to purchase and maintain their own networks and computer systems. In healthcare for example, different initiatives aim at sharing medical images and Personal Health Records (PHR) in between health professionals or hospitals with the help of the cloud. In such an environment, data security (confidentiality, integrity and traceability) is a major issue. In this context that these thesis works, it concerns in particular the securing of Content Based Image Retrieval (CBIR) techniques and machine learning (ML) which are at the heart of diagnostic decision support systems. These techniques make it possible to find similar images to an image not yet interpreted. The goal is to define approaches that can exploit secure externalized data and enable a cloud to provide a diagnostic support. Several mechanisms allow the processing of encrypted data, but most are dependent on interactions between different entities (the user, the cloud or a trusted third party) and must be combined judiciously so as to not leak information. During these three years of thesis, we initially focused on securing an outsourced CBIR system under the constraint of no interaction between the users and the service provider (cloud). In a second step, we have developed a secure machine learning approach based on multilayer perceptron (MLP), whose learning phase can be outsourced in a secure way, the challenge being to ensure the convergence of the MLP. All the data and parameters of the model are encrypted using homomorphic encryption. Because these systems need to use information from multiple sources, each of which outsources its encrypted data under its own key, we are interested in the problem of sharing encrypted data. A problem known by the "Proxy Re-Encryption" (PRE) schemes. In this context, we have proposed the first PRE scheme that allows both the sharing and the processing of encrypted data. We also worked on watermarking scheme over encrypted data in order to trace and verify the integrity of data in this shared environment. The embedded message is accessible whether or not the image is encrypted and provides several services
Ibarrondo, Luis Alberto. "Privacy-preserving biometric recognition systems with advanced cryptographic techniques." Electronic Thesis or Diss., Sorbonne université, 2023. https://theses.hal.science/tel-04058954.
Full textDealing with highly sensitive data, identity management systems must provide adequate privacy protection as they leverage biometrics technology. Wielding Multi-Party Computation (MPC), Homomorphic Encryption (HE) and Functional Encryption (FE), this thesis tackles the design and implementation of practical privacy-preserving biometric systems, from the feature extraction to the matching with enrolled users. This work is consecrated to the design of secure biometric solutions for multiple scenarios, putting special care to balance accuracy and performance with the security guarantees, while improving upon existing works in the domain. We go beyond privacy preservation against semi-honest adversaries by also ensuring correctness facing malicious adversaries. Lastly, we address the leakage of biometric data when revealing the output, a privacy concern often overlooked in the literature. The main contributions of this thesis are: • A new face identification solution built on FE-based private inner product matching mitigating input leakage. • A novel efficient two-party computation protocol, Funshade, to preserve the privacy of biometric thresholded distance metric operations. • An innovative method to perform privacy-preserving biometric identification based on the notion of group testing named Grote. • A new distributed decryption protocol with collaborative masking addressing input leakage, dubbed Colmade. • An honest majority three-party computation protocol, Banners, to perform maliciously secure inference of Binarized Neural Networks. • A HE Python library named Pyfhel, offering a high-level abstraction and low-level functionalities, with applications in teaching
Minelli, Michele. "Fully homomorphic encryption for machine learning." Thesis, Paris Sciences et Lettres (ComUE), 2018. http://www.theses.fr/2018PSLEE056/document.
Full textFully homomorphic encryption enables computation on encrypted data without leaking any information about the underlying data. In short, a party can encrypt some input data, while another party, that does not have access to the decryption key, can blindly perform some computation on this encrypted input. The final result is also encrypted, and it can be recovered only by the party that possesses the secret key. In this thesis, we present new techniques/designs for FHE that are motivated by applications to machine learning, with a particular attention to the problem of homomorphic inference, i.e., the evaluation of already trained cognitive models on encrypted data. First, we propose a novel FHE scheme that is tailored to evaluating neural networks on encrypted inputs. Our scheme achieves complexity that is essentially independent of the number of layers in the network, whereas the efficiency of previously proposed schemes strongly depends on the topology of the network. Second, we present a new technique for achieving circuit privacy for FHE. This allows us to hide the computation that is performed on the encrypted data, as is necessary to protect proprietary machine learning algorithms. Our mechanism incurs very small computational overhead while keeping the same security parameters. Together, these results strengthen the foundations of efficient FHE for machine learning, and pave the way towards practical privacy-preserving deep learning. Finally, we present and implement a protocol based on homomorphic encryption for the problem of private information retrieval, i.e., the scenario where a party wants to query a database held by another party without revealing the query itself
Guellier, Antoine. "Strongly Private Communications in a Homogeneous Network." Thesis, CentraleSupélec, 2017. http://www.theses.fr/2017SUPL0001/document.
Full textWith the development of online communications in the past decades, new privacy concerns have emerged. A lot of research effort have been focusing on concealing relationships in Internet communications. However, most works do not prevent particular network actors from learning the original sender or the intended receiver of a communication. While this level of privacy is satisfactory for the common citizen, it is insufficient in contexts where individuals can be convicted for the mere sending of documents to a third party. This is the case for so-called whistle-blowers, who take personal risks to alert the public of anti-democratic or illegal actions performed by large organisations. In this thesis, we consider a stronger notion of anonymity for peer-to-peer communications on the Internet, and aim at concealing the very fact that users take part in communications. To this end, we deviate from the traditional client-server architecture endorsed by most existing anonymous networks, in favor of a homogeneous, fully distributed architecture in which every user also acts as a relay server, allowing it to conceal its own traffic in the traffic it relays for others. In this setting, we design an Internet overlay inspired from previous works, that also proposes new privacy-enhancing mechanisms, such as the use of relationship pseudonyms for managing identities. We formally prove with state-of-the-art cryptographic proof frameworks that this protocol achieves our privacy goals. Furthermore, a practical study of the protocol shows that it introduces high latency in the delivery of messages, but ensures a high anonymity level even for networks of small size
Ricosset, Thomas. "Signature électronique basée sur les réseaux euclidiens et échantillonnage selon une loi normale discrète." Thesis, Toulouse, INPT, 2018. http://www.theses.fr/2018INPT0106/document.
Full textLattice-based cryptography has generated considerable interest in the last two decades due toattractive features, including conjectured security against quantum attacks, strong securityguarantees from worst-case hardness assumptions and constructions of fully homomorphicencryption schemes. On the other hand, even though it is a crucial part of many lattice-basedschemes, Gaussian sampling is still lagging and continues to limit the effectiveness of this newcryptography. The first goal of this thesis is to improve the efficiency of Gaussian sampling forlattice-based hash-and-sign signature schemes. We propose a non-centered algorithm, with aflexible time-memory tradeoff, as fast as its centered variant for practicable size of precomputedtables. We also use the Rényi divergence to bound the precision requirement to the standarddouble precision. Our second objective is to construct Falcon, a new hash-and-sign signaturescheme, based on the theoretical framework of Gentry, Peikert and Vaikuntanathan for latticebasedsignatures. We instantiate that framework over NTRU lattices with a new trapdoor sampler
Lepoint, Tancrède. "Conception and implémentation de cryptographie à base de réseaux." Phd thesis, Ecole Normale Supérieure de Paris - ENS Paris, 2014. http://tel.archives-ouvertes.fr/tel-01069864.
Full textPérez, Garcia Julio César. "Contribution to security and privacy in the Blockchain-based Internet of Things : Robustness, Reliability, and Scalability." Electronic Thesis or Diss., Avignon, 2023. http://www.theses.fr/2023AVIG0120.
Full textThe Internet of Things (IoT) is a diverse network of objects typically interconnected via the Internet. Given the sensitivity of the information exchanged in IoT applications, it is essential to guarantee security and privacy. This problem is aggravated by the open nature of wireless communications, and the power and computing resource limitations of most IoT devices. Existing IoT security solutions are based on centralized architectures, which raises scalability issues and the single point of failure problem, making them susceptible to denial-of-service attacks and technical failures. Blockchain has emerged as an attractive solution to IoT security and centralization issues. Blockchains replicate a permanent, append-only record of all transactions occurring on a network across multiple devices, keeping them synchronized through a consensus protocol. Blockchain implementation may involve high computational and energy costs for devices. Consequently, solutions based on Fog/Edge computing have been considered in the integration with IoT. However, the cost of Blockchain utilization must be optimized, especially in the consensus protocol, which significantly influences the overall system performance. Permissioned Blockchains align better with the requirements of IoT applications than Permissionless Blockchains, due to their high transaction processing rate and scalability. This is because the consensus nodes, i.e., Validators, are known and predetermined. In existing consensus protocols used in Permissioned Blockchains, the Validators are usually a predefined or randomly selected set of nodes, which affects both system performance and fairness among users. The objective of this work is to propose solutions to improve security and privacy within IoT by integrating Blockchain technology, as well as to maximize fairness levels during consensus. The study is organized into two distinct parts: one addresses critical aspects of IoT security and proposes Blockchain-based solutions, while the other part focuses on optimizing fairness among users during the execution of the consensus algorithm on the Blockchain. We present an authentication mechanism inspired by the µTesla authentication protocol, which uses symmetric keys that form a hashchain and achieves asymmetric properties by unveiling the key used a while later. With this mechanism and the use of the Blockchain to store the keys and facilitate authentication, our proposal ensures robust and efficient authentication of devices, without the need for a trusted third party. In addition, we introduce a Blockchain-based key management system for group communications adapted to IoT contexts. The use of Elliptic Curve Cryptography ensures a low computational cost while enabling secure distribution of group keys. In both security solutions, we provide formal and informal proofs of security under the defined attack model. A performance impact analysis and a comparison with existing solutions are also conducted, showing that the proposed solutions are secure and efficient and can be used in multiple IoT applications. The second part of the work proposes an algorithm to select Validator nodes in Permissioned Blockchains maximizing Social Welfare, using α-Fairness as the objective function. A mathematical model of the problem is developed, and a method for finding the solution in a distributed manner is proposed, employing metaheuristic Evolutionary algorithms and a Searchspace partitioning strategy. The security of the proposed algorithm and the quality of the solutions obtained are analyzed. As a result of this work, two security protocols for IoT based on Blockchain are introduced, along with a distributed algorithm for maximizing Social Welfare among users in a Permissioned Blockchain network
Bozdemir, Beyza. "Privacy-preserving machine learning techniques." Electronic Thesis or Diss., Sorbonne université, 2021. http://www.theses.fr/2021SORUS323.
Full textMachine Learning as a Service (MLaaS) refers to a service that enables companies to delegate their machine learning tasks to single or multiple untrusted but powerful third parties, namely cloud servers. Thanks to MLaaS, the need for computational resources and domain expertise required to execute machine learning techniques is significantly reduced. Nevertheless, companies face increasing challenges with ensuring data privacy guarantees and compliance with the data protection regulations. Executing machine learning tasks over sensitive data requires the design of privacy-preserving protocols for machine learning techniques.In this thesis, we aim to design such protocols for MLaaS and study three machine learning techniques: Neural network classification, trajectory clustering, and data aggregation under privacy protection. In our solutions, our goal is to guarantee data privacy while keeping an acceptable level of performance and accuracy/quality evaluation when executing the privacy-preserving variants of these machine learning techniques. In order to ensure data privacy, we employ several advanced cryptographic techniques: Secure two-party computation, homomorphic encryption, homomorphic proxy re-encryption, multi-key homomorphic encryption, and threshold homomorphic encryption. We have implemented our privacy-preserving protocols and studied the trade-off between privacy, efficiency, and accuracy/quality evaluation for each of them
Mkhinini, Asma. "Implantation matérielle de chiffrements homomorphiques." Thesis, Université Grenoble Alpes (ComUE), 2017. http://www.theses.fr/2017GREAT092/document.
Full textOne of the most significant advances in cryptography in recent years is certainly the introduction of the first fully homomorphic encryption scheme by Craig Gentry. This type of cryptosystem allows performing arbitrarily complex computations on encrypted data, without decrypting it. This particularity allows meeting the requirements of security and data protection, for example in the context of the rapid development of cloud computing and the internet of things. The algorithms implemented are currently very time-consuming, and most of them are implemented in software. This thesis deals with the hardware acceleration of homomorphic encryption schemes. A study of the primitives used by these schemes and the possibility of their hardware implementation is presented. Then, a new approach allowing the implementation of the two most expensive functions is proposed. Our approach exploits the high-level synthesis. It has the particularity of being very flexible and generic and makes possible to process operands of arbitrary large sizes. This feature allows it to target a wide range of applications and to apply optimizations such as batching. The performance of our co-design was evaluated on one of the most recent and efficient homomorphic cryptosystems. It can be adapted to other homomorphic schemes or, more generally, in the context of lattice-based cryptography
Vial, prado Francisco. "Contributions to design and analysis of Fully Homomorphic Encryption schemes." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLV107/document.
Full textFully Homomorphic Encryption schemes allow public processing of encrypted data. Since the groundbreaking discovery of the first FHE scheme in 2009 by Craig Gentry, an impressive amount of research has been conducted to improve efficiency, achieve new levels of security, and describe real applications and connections to other areas of cryptography. In this Dissertation, we first give a detailed account on research these past years. Our contributions include a key-recovery attack on the ideal lattices FHE scheme and a new conception of hierarchic encryption, avoiding at some extent betrayal between users while maintaining the flexibility of FHE. We also describe some implementations. This research was done in the Laboratoire de Mathématiques de Versailles, under supervision of Prof. Louis Goubin
Decouchant, Jérémie. "Collusions and Privacy in Rational-Resilient Gossip." Thesis, Université Grenoble Alpes (ComUE), 2015. http://www.theses.fr/2015GREAM034/document.
Full textGossip-based content dissemination protocols are a scalable and cheap alternative to centralised content sharing systems. However, it is well known that these protocols suffer from rational nodes, i.e., nodes that aim at downloading the content without contributing their fair share to the system. While the problem of rational nodes that act individually has been well addressed in the literature, textit{colluding} rational nodes is still an open issue. In addition, previous rational-resilient gossip-based solutions require nodes to log their interactions with others, and disclose the content of their logs, which may disclose sensitive information. Nowadays, a consensus exists on the necessity of reinforcing the control of users on their personal information. Nonetheless, to the best of our knowledge no privacy-preserving rational-resilient gossip-based content dissemination system exists.The contributions of this thesis are twofold.First, we present AcTinG, a protocol that prevents rational collusions in gossip-based content dissemination protocols, while guaranteeing zero false positive accusations. AcTing makes nodes maintain secure logs and mutually check each others' correctness thanks to verifiable but non predictable audits. As a consequence of its design, it is shown to be a Nash-equilibrium. A performance evaluation shows that AcTinG is able to deliver all messages despite the presence of colluders, and exhibits similar scalability properties as standard gossip-based dissemination protocols.Second, we describe PAG, the first accountable and privacy-preserving gossip protocol. PAG builds on a monitoring infrastructure, and homomorphic cryptographic procedures to provide privacy to nodes while making sure that nodes forward the content they receive. The theoretical evaluation of PAG shows that breaking the privacy of interactions is difficult, even in presence of a global and active opponent. We assess this protocol both in terms of privacy and performance using a deployment performed on a cluster of machines, simulations involving up to a million of nodes, and theoretical proofs. The bandwidth overhead is much lower than existing anonymous communication protocols, while still being practical in terms of CPU usage