Abstract: How can we run large AI models on untrusted hardware without revealing the inputs or the model? We present MOSAIC, a scalable cryptographic protocol for outsourcing AI computation in large-scale data centers. A small trusted client, bootstrapped from secure boot and trusted execution technologies, holds the private input and model weights, and outsources most
Abstract: Response and prevention of cyber incidence requires analysis of information from multiple sources. Endpoint monitoring solutions continuously record system-level activities as audit logs, providing deep visibility into security incidents. Yet before designing increasingly sophisticated solutions, it is essential to understand what system provenance can reveal, how its information can be abstracted, and where intelligent
Abstract: The security of RSA is generally understood to be based on the complexity of factoring, and key size parameters are extrapolated from the general number field sieve (GNFS). However, this may not accurately represent RSA security in practical scenarios. An under-appreciated 2007 algorithm of Joux, Naccache, and Thomé allows an attacker to forge RSA
Abstract: Autonomous systems can complete individual security tasks while producing behaviour that is poorly suited to a red teaming exercise. This becomes particularly apparent when several agents are operating concurrently, as local decisions can accumulate to form a pattern of activity that is inconsistent with the environment or the intended adversary. Therefore, behavioural stealth raises
Abstract: A modern computer system, be a scale-up server or a phone, is a complex mixture of heterogeneous cores, firmware images, management processors, etc. The “de facto” operating system of such a machine, therefore, is itself a large ad-hoc collection of components (of which Linux or MacOS is only one) that have no explicit correctness
Abstract: Trusted computing mechanisms such as trusted execution environments, and derived services such as remote attestation, and proofs of execution are increasingly important for securing embedded systems. However, many of these mechanisms rely on atomic execution and interrupt suppression to preserve integrity guarantees, directly conflicting with the timing and availability requirements of real-time embedded systems
Abstract: Confidential Virtual Machines (CVMs) are the cornerstone of the modern push for Private AI and the future “default” for protecting data in use, however, the transition from architectural specification to secure deployment is fraught with challenges. While SEV-SNP and TDX offer a new paradigm for data-in-use protection, their security properties are not static. As
Abstract: The current geopolitical situation is characterized by growing dependence on digital infrastructures and increasing international tensions. For Switzerland, the Domain Name System (DNS) represents a crucial component in ensuring digital sovereignty and the protection of national interests. As a neutral state, Switzerland is particularly committed to maintaining its digital infrastructure in an independent, resilient,
Abstract: Abstract: Layer 2 rollups are rapidly absorbing DeFi activity, securing over $40 billion and accounting for nearly half of Ethereum’s DEX volume by Q1 2025, yet their MEV dynamics remain understudied. We address this gap by defining and quantifying optimistic MEV, a form of speculative, on-chain MEV whose detection and execution logic reside largely
Abstract: Cryptographic software is currently facing two major challenges. First, upgrading our cryptographic infrastructure to post-quantum primitives is probably the largest and most demanding cryptographic migration effort ever. New software needs to be written, optimized for different platforms, extensively tested and audited, and eventually integrated into protocols and systems. Second, it becomes increasingly clear that