With the pervasive digitalization of modern life, we benefit from efficient access to information and services. Yet, this digitalization poses severe privacy challenges, especially for special-needs individuals. Beyond being a fundamental human right, privacy is crucial for roles sensitive in nature, including investigative journalists exposing corruption and humanitarian organizations supporting refugees or survivors of violence. This thesis leverages privacy-enhancing technologies to mitigate the risks of digitalization while retaining its advantages. Recent breakthroughs in cryptography, such as fully homomorphic encryption and secure multiparty computation, provide robust tools for privacy. However, there is still no silver bullet solution that can achieve efficient privacy out of the box. We observe that there often is a gap between theoretical cryptographic solutions and real-world problems. Identifying and bridging these gaps enables us to design pragmatic privacy-enhancing technologies tailored for real-world deployment. In this thesis, we identify and solve four real-world problems. We first present the problem of searching sensitive documents among a network of investigative journalists. In collaboration with the International Consortium of Investigative Journalists, we design a decentralized peer-to-peer privacy-preserving search engine called DatashareNetwork. Our solution enables journalists to find colleagues who have relevant documents for their topic of investigation and anonymously discuss the possibility of collaboration. We develop a prototype of DatashareNetwork and demonstrate that it scales to thousands of journalists and millions of documents. We introduce a new class of problems called private collection matching in which a client aims to determine whether a collection of sets owned by a server matches their interests such as searching confidential chemical compound databases. We design a framework based on fully homomorphic encryption to solve these problems. Our solution, takes the data minimization principle to the maximum and shows the possibility of satisfying clients' needs by only revealing a single bit. We evaluate our framework and show that it significantly improves the latency, client computation cost, and communication cost with respect to generic solutions that offer the same privacy guarantee. We examine the problem of preventing double registration in humanitarian aid distribution with a focus on the needs of the International Committee of Red Cross. In response, we design Janus, a privacy-preserving biometric deduplication system that is compatible with fingerprints, irises, and face recognition; and supports both biometric alignment and fusion. We design and develop three instantiations of Janus based on secure multiparty computation, somewhat homomorphic encryption, and trusted execution environments. We evaluate Janus to show it satisfies the privacy, accuracy, and performance needs of humanitarian organizations. Finally, we study the problem of detecting insecure ciphers in aircraft communication at scale. We design and develop a decision support system that helps human analysts to detect new ciphertexts in aircraft communication. We evaluate our system by applying it to real-world data and asking our analyst to use our support system to find new ciphers. Our analysis led to uncovering of 9 previously unknown (and potentially insecure) ciphers which we disclose to various stakeholders.