Cryptography and Security Papers of the Week

We introduce the Red-Green Sequence (RGS) watermark, a private and computationally efficient watermark that achieves accurate frequency estimation at high privacy levels across various models, sample sizes and generation entropies in settings that reflect realistic deployment scenarios.

Securing this infrastruc-ture requires cryptographic integrity and confidential-ity under unprecedented constraints: ultra-low microsec-ond latency deadlines to avert catastrophic arcing dam-age (< 50 µs), microcontrollers with severely constrained SRAM and Flash budgets, and an electromagnetically violent environment where high-power spark transients (dv/dt > 10 kV/µs, di/dt > 109 A/s) induce severe ra-diated electromagnetic interference (EMI). In this pa-per, we present an exhaustive empirical performance and security evaluation of five premier National Insti-tute of Standards and Technology (NIST) Lightweight Cryptography (LWC) authenticated encryption with as-sociated data (AEAD) ciphers—Ascon-128, Ascon-128a, TinyJAMBU-128, Photon-Beetle-AEAD, and Xoodyak—alongside GIFT-COFB and the legacy enterprise standard AES-128-GCM.

We design and implement a system for bilateral protection of private LoRA inference using Intel Trust Domain Extensions (TDX), connecting approved application identity to actual artifact consumption and adapter activation.

This paper presents KVGuard, the first systematic memory-safety abstraction for KV cache management in LLM serving systems.

In this work, we systematically study the risk of user-profile inference from MCP tool-call traces, where observers infer user attributes from protocol-valid tool interactions without access to user conversations.

In this paper, we propose an adversarial self-evolution framework that integrates bi-level optimization between the attacker's strategy evolution and the defender's adaptive mitigation, to expose the vulnerabilities of search agents.

Guided by this model, we develop a Multi-Agent Reinforcement Learning (MARL) detector coupled with signal repair at local control centers, designed to mitigate distributional data drift through complementary feature-based detection and clean-reference-calibrated repair while meeting latency budgets.

This paper presents TrajectoryGuard , a personalized and privacy-preserving federated learning framework for secure and efficient mobility data mining.

In the security vulnerability reasoning study, probes and activation ablations provide evidence of internal representation and causal influence, but a monitor with access to the reference answer does not outperform the majority-class baseline in a held-out test.

This work not only provides a theoretical basis for the design of highly efficient luminescent CDs but also expands their potential in intelligent materials and photonic applications.

We propose the Open Inference Market (OIM), which replaces the missing asymmetry with an economic mechanism.

Existing defenses can block the attack yet impair the model's use of that information, leaving a gap between attack defense and task preservation. We propose ERASE, an on-policy self-distillation method that constructs its reference by erasing the injected region in place.

We introduce germline execution, which limits information disclosure without inspecting or trusting tool implementations.

We show that visual perturbations can instead alter the physical action while keeping the generated rationale word-for-word identical.