Certified in Theory, Broken in Practice: Assumption Gaps in Cryptographic Model Certification

Manipulation-Proof Oblivious Audits against Deceptive Model Providers

Audits have emerged as a critical instrument for algorithmic governance, providing a mechanism for external scrutiny and governance of machine learning models. However, ensuring the integrity of such assessments remains a challenging issue. For instance in regulatory contexts, audits are typically declared or easily detected, thus enabling model providers to manipulate the process, whether intentionally or inadvertently. This vulnerability is particularly acute in the context of fairness evaluations, in which providers can often infer sensitive attributes and strategically equalize allocation rates between groups to satisfy fairness metrics. In this paper, we introduce a novel audit protocol designed to significantly increase the post-audit detectability of such manipulations by enabling the auditor to query the model in an oblivious manner. Our approach leverages a Private Information Retrieval mechanism to require the provider to label a large set of instances, while preventing it from knowing which subset will ultimately be used for the audit. The protocol is efficient, imposes minimal overhead on the auditor, and requires no modification to the audited model, its training procedure, or its inference pipeline. We provide theoretical guarantees showing that, under this protocol, a provider attempting to hide unfairness must falsify a significantly larger number of responses, thereby increasing both the difficulty and the likelihood of detection of manipulation. Experimental results across representative audit scenarios confirm the effectiveness and practicality of our approach.

SoK: Verifiable Integrity Claims for Privacy-Preserving Federated Learning

This SoK model federated learning as an append-only transcript of submissions, admissions, aggregation, and finalization events, and formalize verifiability as a collection of integrity claims issued by clients and the aggregator, and checked by different verifier classes.

CT-Monitor++: Improving the Certificate Transparency Ecosystem with Cryptographic and Security Checks

: Digital certificates are one of the main tools to provide data and public key authentication, but their use in a Public Key Infrastructure (PKI) relies on the trustworthiness of each individual Certification Authority (CA), which may be compromised or act maliciously. Certificate Transparency (CT) mitigates this risk by requiring all issued certificates to be publicly logged, enabling the detection of unauthorized certificates. In this work, we present the design of a CT monitor that goes beyond traditional domain-based tracking by performing systematic security analysis on all certificates recorded in public CT logs. Our monitor executes a comprehensive set of security checks, inspecting certificates for cryptographic weaknesses, syntactic inconsistencies, and potentially malicious contents, providing real-time insights into the security of the Web PKI ecosystem. We implement and evaluate our monitor on live CT log streams, analyzing its computational overhead, memory requirements, and scalability as certificate volumes increase. We quantify the resources needed to perform these analyses in real time and show that, with the proposed optimizations, our monitor is capable of keeping up with the current certificate issuance rates while performing all security checks.

Migrating to Hybrid Cryptography in Practice: The TutaCrypt Protocol and Its Security

In this work, we study the hybrid key establishment protocol TutaCrypt as a concrete case of post-quantum cryptographic migration in practice. The protocol was developed by the end-to-end encrypted email provider Tuta and has been deployed to more than ten million users worldwide. We present the protocol in a form that enables rigorous cryptographic analysis and define two Bellare–Rogaway-style security models that precisely characterize the provided security guarantees. The two models capture that the security properties achieved in the pre-quantum setting are slightly stronger than in the post-quantum setting. We then give reduction-based security proofs that clarify under which assumptions the construction achieves security, and how its guarantees degrade if either the classical or the post-quantum component is compromised. Our results illustrate how formally grounded cryptographic models can capture real-world migration strategies and hybrid deployments. Such analyses help to understand the security properties of deployed cryptographic systems and help move cryptographic migration from best practice toward principled, verifiable design.

On the Privacy-Preserving Capabilities of PAVE Specifications in Learnware ∗

Privacy-Preserving Conformance Checking Using Quantum-Safe Fully Homomorphic Encryption

Conformance checking is a fundamental task in process mining that evaluates how well the observed executions recorded in event logs conform to a given process model. This task enables the identification of deviations, inefficiencies, and bottlenecks in real-world business processes, thereby supporting process improvement and compliance analysis. However, when business processes are deployed in untrusted environments, outsourcing both event logs and process models introduces significant privacy risks, as sensitive operational information may be exposed to external service providers. Existing privacy-preserving conformance checking (PPCC) approaches remain limited, as they primarily rely on partial anonymization, secure multiparty computation, or lightweight encryption mechanisms, without enabling computation directly over encrypted data. This paper introduces a novel PPCC method based on post-quantum fully homomorphic encryption (FHE). The proposed method performs token-based replay fitness evaluation directly over encrypted process models and encrypted event logs. Specifically, the method incorporates a privacy-preserving token-based replay mechanism that homomorphically evaluates encrypted Petri net models and encrypted event logs, enabling the fitness value to be obtained while preserving data confidentiality throughout the computation process. To the best of our knowledge, this is the first method that enables token-based replay fitness computation entirely in the encrypted domain using post-quantum FHE. An experimental evaluation was conducted using synthetic and real-life event logs to analyze the proposed method in terms of utility preservation, computational performance and scalability. The results demonstrate that the encrypted-domain fitness computation produces results equivalent to those obtained in the plaintext setting, while introducing manageable computational overhead. By combining formal conformance checking techniques with quantum-resistant cryptography, the proposed method enables privacy-preserving fitness analysis under outsourced and untrusted environments.