La transmission de données multimédia sur les réseaux sécurisés a une croissance exponentielle grâce aux progrès scientifique dans les technologies de l'information et de la communication. La sécurité des données dans certaines applications comme le stockage sécurisé, l'authentification, la protection des droits d'auteurs, la communication militaire ou la visioconférence confidentielles, nécessitent de nouvelles stratégies en matière de transmission sécurisée. Deux techniques sont couramment utilisées pour la transmission sécurisée de données visuelles, à savoir : la cryptographie et la stéganographie. La cryptographie sécurise les données en utilisant des clés secrètes afin de rendre les données illisibles, la stéganographie, elle, vise à insérer des données cruciales dans des signaux porteurs anodins.De plus, pour la confiance mutuelle et les systèmes distribués, le partage sécurisé de ressources est souvent une garantie suffisante pour les applications de communication. L'objectif principal de cette thèse est de réaliser une protection des données visuelles, en particulier les images numériques, par le biais des techniques modernes de cryptographie. Dans ce contexte, deux objectifs de recherche ont été développés durant ces travaux de thèse.La première partie de notre travail se concentre sur la sécurité des images numériques dans un environnement partagé. Ensuite, la deuxième partie porte sur l'intégrité des données visuelles pendant une transmission sécurisée.Nous avons proposé un nouveau schéma de partage des images qui exploite les propriétés d'addition et de multiplication homomorphique de deux crypto systèmes à clé publique largement utilisés : les algorithmes RSA et Paillier. Dans les schémas traditionnels de partage sécurisé, le ``dealer'' partitionne le secret en parties et le distribue à chacun des autres acteurs. Ainsi, aucun des acteurs impliqués ne participe à la création du partage sécurisé, mais il est toujours possible que le ``dealer'' transmette des données malveillantes. Au contraire, l'approche proposée utilise le système de partage de secret d'une manière qui limite l'influence du ‘‘dealer'' sur le protocole en permettant à chaque acteur de participer.La deuxième partie de ces travaux de thèse met l'accent sur l'intégrité des données visuelles lors de la transmission. L'intégrité des données signifie que les données gardent leurs structures complètes au cours d'une opération numérique comme le stockage, le transfert ou la récupération. Le changement d'un seul bit de données cryptées peut avoir un impact catastrophique sur les données décryptées. Nous abordons le problème de correction d'erreurs dans les images cryptées en utilisant le chiffrement à clé symétrique AES (Advanced Encryptions Standard) suivant différents modes. Trois mesures sont proposées afin d'exploiter les statistiques locales des données visuelles et l'algorithme de chiffrement, dans l'objectif de corriger les erreurs efficacement. / Due to the advancements in the information and communication technologies, the transmission of multimedia data over secure or insecure communication channels has increased exponentially. The security of data in applications like safe storage, authentications, copyright protection,remote military image communication or confidential video-conferencing require new strategies for secure transmission. Two techniques are commonly used for the secure transmission of visual data, i.e. cryptography and steganography. Cryptography achieves security by using secret keysto make the data illegible while steganography aims to hide the data in some innocent carrier signal. For shared trust and distributed environment, secret sharing schemes provide sufficient security in various communication applications. The principal objective of this thesis is to achieveprotection of visual data especially images through modern cryptographic techniques. In this context, the focus of the work in perspective, is twofolded. The first part of our work focuses on the security of image data in shared environment while the second part focuses on the integrity ofimage data in the encrypted domain during transmission.We proposed a new sharing scheme for images which exploits the additive and multiplicative homomorphic properties of two well known public key cryptosystems, namely, the RSA and the Paillier. In traditional secret sharing schemes, the dealer partitions the secret into shares and distributethe shares to each of the player. Thus, none of the involved players participate in the creation of the shared secret and there is always a possibilitythat the dealer can cheat some player. On the contrary, the proposed approach employs the secret sharing scheme in a way that limits the influence of the dealer over the protocol by allowing each player to participate. The second part of our thesis emphasizes on the integrity of visual data during transmission. Data integrity means that the data have its complete structure during any operation like storage, transfer or retrieval. A single bit change in encrypted data can have catastrophic impact over the decrypted data. We address the problem of error correction in images encrypted using symmetric key cryptosystem of the Advanced Encryption Standard (AES) algorithm. Three methods are proposed to exploit the local statistics of the visual data and the encryption algorithm to successfully correct the errors.

The development of innovative applications of Artificial Intelligence (AI) is inseparable from the sharing of public data. However, as people strengthen their awareness of the protection of personal data privacy, it is more and more difficult to collect data from multiple data sources and there is also a risk of leakage in unified data management. But neural networks need a lot of data for model learning and analysis. Federated learning (FL) can solve the above difficulties. It allows the server to learn from the local data of multiple clients without collecting them. This thesis mainly deploys FL on the Raspberry Pi (RPi) and achieves federated averaging (FedAvg) as aggregation method. First in the simulation, we compare the difference between FL and centralized learning (CL). Then we build a reliable communication system based on socket on testbed and implement FL on those devices. In addition, the Paillier encryption algorithm is configured for the communication in FL to avoid model parameters being exposed to public network directly. In other words, the project builds a complete and secure FL system based on hardware. / Utvecklingen av innovativa applikationer för artificiell intelligens (AI) är oskiljaktig från delning av offentlig data. Men eftersom människor stärker sin medvetenhet om skyddet av personuppgiftsskydd är det allt svårare att samla in data från flera datakällor och det finns också risk för läckage i enhetlig datahantering. Men neurala nätverk behöver mycket data för modellinlärning och analys. Federated learning (FL) kan lösa ovanstående svårigheter. Det gör det möjligt för servern att lära av lokala klientdata utan att samla in dem. Denna avhandling använder huvudsakligen FL på Raspberry Pi (RPi) och uppnår federerad genomsnitt (FedAvg) som aggregeringsmetod. Först i simuleringen jämför vi skillnaden mellan FL och CL. Sedan bygger vi ett pålitligt kommunikationssystem baserat på uttag på testbädd och implementerar FL på dessa enheter. Dessutom är Paillier -krypteringsalgoritmen konfigurerad för kommunikation i FL för att undvika att modellparametrar exponeras för det offentliga nätverket direkt. Med andra ord bygger projektet ett komplett och säkert FL -system baserat på hårdvara.

Dissertação (mestrado)—Universidade de Brasília, Faculdade de Tecnologia, Departamento de Engenharia Elétrica, 2016. / Submitted by Fernanda Percia França (fernandafranca@bce.unb.br) on 2017-03-08T21:22:12Z No. of bitstreams: 1 2016_VitorVenezaQuimasMacedo.pdf: 1728491 bytes, checksum: e5d32d334677321f2b5ea3b5c03b0b41 (MD5) / Approved for entry into archive by Raquel Viana(raquelviana@bce.unb.br) on 2017-03-24T19:03:33Z (GMT) No. of bitstreams: 1 2016_VitorVenezaQuimasMacedo.pdf: 1728491 bytes, checksum: e5d32d334677321f2b5ea3b5c03b0b41 (MD5) / Made available in DSpace on 2017-03-24T19:03:33Z (GMT). No. of bitstreams: 1 2016_VitorVenezaQuimasMacedo.pdf: 1728491 bytes, checksum: e5d32d334677321f2b5ea3b5c03b0b41 (MD5) / O objetivo do presente trabalho é provar a segurança e eficiência de um protocolo que implementa a interseção privada de conjuntos (Private Set Intersection - PSI) entre dois participantes, Alice e Bob. Neste protocolo Bob possui um conjunto de elementos e deseja saber se este pertence ao conjunto de pontos de Alice, sem, no entanto, Alice saber nada sobre o resultado e Bob não saber nada além da interseção e o tamanho do conjunto de Alice. Para tanto é utilizada, numa primeira abordagem, a avaliação inconsciente de polinômio (ObliviousPolynomialEvaluation) com o uso de uma Terceira Parte Confiável (TPC), conseguindo obter dessa forma a segurança incondicional, em ambientes estáticos, contra adversários ativos. Na sequência é apresentada uma segunda versão do protocolo onde Alice e Bob simulam o papel da TPC, através do uso do algoritmo criptográfico homomórfico de Paillier, porém neste caso obtendo segurança computacional, em ambientes estáticos, contra adversário passivos. Por último é apresentada uma terceira versão do protocolo onde Alice e Bob, sem uma TPC, em uma única fase, usando o algoritmo criptográfico homomórfico de Paillier, em que Bob consegue testar um conjunto com mais do que apenas um elemento, obtendo novamente segurança computacional, em ambientes estáticos, contra adversário passivos Considerando as pesquisas bibliográficas realizadas, este é o primeiro protocolo que implementa a interseção privada de conjuntos provado incondicionalmente seguro no modelo Composto Universalmente (UniversallyComposable - UC). / The present paper goal is to prove the security and efficiency of a protocol that implements private set intersection – PSI between two players, Alice and Bob. Bob possess a set of elements and wishes to know if any of these elements belongs also in Alice set of elements, without, however, Alice knowing anything about the intersection and Bob knowing nothing besides the intersection and the size of Alice set. To conquer this objective, it is used, at a first approach, the oblivious polynomial evaluation using a Trusted Third Party (TTP), which achieves unconditional security, in static environments, against active adversaries. Moreover, it is presented a second version of the protocol where Alice and Bob simulate the TTP role through the use of Paillier homomorphic cryptography algorithm, but in this scenario it is achieved computational security, in static environments, against passive adversaries. At last, it is presented a third version of the protocol, without a TTP, still using Paillier homomorphic cryptography algorithm, where it is possible to test Bob set with more than one element, with the same security level as of the second version. Considering the previous bibliographic research, this is the first protocol that implements private set intersection proved unconditional secure in the Universally Composable - UC -model.

區塊鏈有著不可否認性、可追溯性以及共識性等特點，所有的交易內容都會完整的被記錄在區塊鏈上，基於上述幾項特性，我們利用區塊鏈來記錄公開資訊，將私密資料經由分散式秘密共享後再加密存放於智能合約中。其中，智能合約是一個能將交易狀態和交易狀態內嵌於區塊鏈上的應用，透過智能合約作為媒介，我們能夠將加密後的私密資料完整的存放於區塊鏈上，最後經由區塊鏈網路上的節點驗證後，達到資料正確性驗證的目的。 本研究分析現有的電子投票以及電子投標等應用的系統架構後，發現兩者皆存在著可信賴的第三方進行開票及開標的角色，且驗證流程繁瑣，無法提供一個便利性的投票與投標流程。此外，上述兩種應用皆須滿足機密性、不可否認性、匿名性以及可驗證性等安全性質，若能結合區塊鏈與智能合約於上述應用中，將可提升資料的可驗證性以及降低成本的負擔，對參與應用的人而言也能達到公開透明的需求。 因此，本文提出一個分散式架構下的電子投票與投標機制，結合區塊鏈以及智能合約的優點與技術，讓所有參與投票的選民、投標的廠商共同參與驗證與計算，並加強參與者的匿名性、資料傳輸的隱私性、開票與開標階段資料的可信賴性以及可驗證性。 / With the rise of blockchain technology, the core concept of decentralization has gradually drawn attention. In this context, the main objective of this study is to realize more convenient and secure electronic applications with the use of blockchain technology. This research is aimed to design a distributed e-voting and e-bidding system. The core idea is to combine the blockchain technology with secret sharing scheme and homomorphic encryption in order to realize the distributed e-voting and e-bidding application without a trusted third party. The system allows voters to participate in opening phase. It provides a public and transparent process while protecting the anonymity of voter’s and vendor’s identity, the privacy of data transmission and verifiability of data during the opening phase.