Cryptography is used everywhere in our society, from simple internet searches to providing security and privacy for our banking systems. A key research area in this field is secure function evaluation, where multiple parties compute a function of their inputs, without revealing them to each other. However, as most cryptographic tasks, current implementations often rely on computational assumptions, which may become insecure in the future. Researchers have explored quantum communications to eliminate such computational assumptions since the 80s, but still challenges remain. For instance, many theoretical protocols using quantum communications rely on perfect single-photon sources. However, in practice, photon sources have a small probability of emitting multiple photons in one pulse, potentially leaking information to a dishonest party. Furthermore, a dishonest party could have secretly tampered the measurement device of the other party and potentially introduced a backdoor in it. The so-called “measurement device independent” setting has then been described, to provide security against this kind of attacks.
In this work, we investigate the possibility of achieving two secure function evaluation primitives - oblivious transfer and bit commitment - in the measurement device independent setting, using imperfect single-photon sources. We already know that under this setting, bit commitment is possible to achieve but oblivious transfer seems hard to achieve. It is still an open question if there exist a protocol to achieve oblivious transfer under this setting. If not, it would be interesting to find new achievable primitives other than oblivious transfer. Here, we study the existing protocols and provide some results towards answering these questions. We start by extending an existing result on the difficulty of achieving oblivious transfer by proving that even a weaker version of this primitive is difficult to achieve under such setting. We also shortly explore the possibility of using an existing reduction from oblivious transfer to bit commitment to see whether we can achieve oblivious transfer from the existing bit commitment protocol. Finally, we show that an existing bit commitment protocol can perform at a better rate by giving a slightly modified security proof.

While securing data-in-use was assured by well-known encryption algorithms, the industry shifted towards trusting hardware manufacturers in exchange for efficiency speedups through Trusted Execution Environments. However, there are many technologies to choose from, each with its own design and trade-offs. Additionally, no work was conducted to systematically compare Trusted Execution Environments side-by-side with privacy-preserving techniques. Therefore, this literature review analyzes, in the first part, Intel's SGX, Keystone, Intel's TDX, and AMD's SEV from four angles that are strongly tied to data-in-use protection (functionality, efficiency, security, and usability). We observed that even though complex and inherently suffering to hardware-related attacks, TEEs offer a great option for confidential computing. Lastly, this research compares these state-of-the-art technologies to four other privacy-preserving techniques (Fully Homomorphic Encryption, Secure Multi-Party Computation, Oblivious RAM, and Structured Encryption) by drawing common properties and displaying them in equal use cases, showing that TEEs are a great choice for many use cases, but with stronger security issues.

With the ever-growing cloud capacity in our society, being able to harness that immense computational power while keeping our data private has become a strong point of interest. This can be achieved with Structured Encryption, which allows a data structure to be encrypted and stored remotely, but still queryable by the client owning
the encryption key. This report is a literature review of the field of Structured Encryption (StE). We analyze the state-of-the-art technologies and their characteristics on the following aspects: security, efficiency, functionality and usability, and discuss their capabilities and limitations. We then compare StE schemes with other promising technologies in the area of computation on encrypted data: Fully Homomorphic Encryption, Oblivious RAM, Secure Multiparty Computation and Trusted Execution Environments.

Outsourcing data to the cloud can pose serious security threats due to an attacker observing the data access patterns, even though data is encrypted. Ideally, confidentiality should not depend on the server being a trusted party. Oblivious Random Access Machines are tackling this problem by obfuscating data access patterns and ensuring obliviousness of the server towards the data. Thus, this paper studies the evolution of Oblivious Random Access Machines and highlights the steps taken so far to discover a more practical algorithm for hiding data access patterns on an un trusted server. In addition, other approaches for privacy-preserving computation, such as Homomorphic Encryption, Structured Encryption, Multi Party Computation and Trusted Execution Environments are discussed and contrasted to ORAM to assess the costs and benefits they come with, but also the trade-offs in security and usability.

Secure multi-party computation (MPC) allows parties to compute on their secret inputs, without revealing them to each other. As an area of theoretical interest, many MPC protocol have been developed in the last four decades. They each present different characteristics and are classified under distinct categories depending on their generality, security assumptions, and functionality. More recently, MPC has also become an area of practical interest due to optimizations in performance of the protocols. In this paper, we compare MPC protocols and other techniques for computing on encrypted data, considering how their properties affect security, efficiency, usability, and functionality. We show that there is a trade-off between security and efficiency when different adversarial models are used, as well as a trade-off between efficiency and flexibility in specialized protocols.

In the digital economy, individual data sovereignty is a requirement for sustainable and secure development. With an increase in the number of outsourced computations, due to the adoption of web services, artificial intelligence and research-based innovation, privacy became one of the main concerns for individuals and third parties alike. Fully Homomorphic Encryption (FHE) addresses this gap by enabling secure computations on encrypted data. This paper reviews and presents the FHE literature in a structural manner, covering the topics of mathematical constructions, security, efficiency and functionality. In doing so, it highlights the current state-of-the-art techniques and provides a systemic comparison with other privacy enhancing mechanisms. In addition, it discusses its applicability in the practical domain, such as Confidential Machine Learning, Medical Data Analysis and Recommender Systems, mentioning potential impediments and their solutions for mass adoption.

As technology for implementing quantum networks advances, the challenge of efficiently managing entanglement generation under resource constraints becomes critical. In this thesis, a systematic methodology for selecting scheduling strategies in a centralised quantum network is proposed, and applied to a specific case for a quantum hub network utilising an entanglement generation switch. Using the NetSquid simulator, the first in, first out, on-demand, MaxWeight, earliest deadline first, earliest feasible deadline first and round-robin scheduling algorithms are evaluated across a wide range of metrics compiled in this thesis.
These metrics include measures of throughput, fairness, responsiveness, demand completion, and resource utilisation, among others. Varying network load conditions are simulated and two application-level use cases, quantum key distribution and blind quantum computing, are considered. The results offer detailed insight into how each scheduler performs under different demand patterns and operational contexts.
Based on these results, the earliest deadline first performs better in most metrics than the other schedulers within the context of this thesis. Additionally, the results indicate that classical optimality of a scheduler does not always translate to superior performance in quantum network scenarios. The findings presented and the framework described here can provide practical guidance for network operators seeking to balance multiple performance goals in near-term quantum networks.