Deep-learning models are increasingly deployed in image recognition and intelligent decision-making, making model copyright verification important for protecting valuable digital assets. Existing structure- or parameter-based comparison methods are difficult to apply when only black-box access to a target model is available, and ordinary test samples often fail to reveal decision-boundary differences. This paper proposes a black-box copyright verification method based on adaptive boundary fingerprints. The method generates candidate boundary samples using the Fast Gradient Sign Method (FGSM), selects fingerprints under source correctness, margin, class balance, and diversity constraints, and measures source–target behavioral similarity by fusing label consistency, probability-distribution similarity, and boundary transition consistency. A decision threshold is then calibrated from the background distribution of independent models. Experiments on CIFAR-100 and Tiny-ImageNet show that independent models remain below calibrated thresholds, while most derived models obtain higher scores. On CIFAR-100, the method achieves a 91.7% detection rate and a 0.0% false-positive rate, with Area Under the Receiver Operating Characteristic Curve (AUROC) and Area Under the Precision-Recall Curve (AUPR) of 0.9750 and 0.9643, respectively. These results indicate that adaptive boundary fingerprints provide effective evidence for black-box model copyright verification.
As security authentication issues continue to arise in future wireless communication networks, researchers are working hard to further improve authentication techniques. Recently, physical layer authentication (PLA) has received widespread attention for its lightweight nature compared to traditional encryption methods based on keys and blockchain. However, the existing PLA mechanisms based on a fixed decision threshold have low reliability in dynamic environments. Moreover, PLA solutions are typically based on binary authentication, and these binary-type schemes cannot provide different levels of access control. To address these challenges, this article introduces the concept of hierarchical security authentication, aiming to achieve multi-level secure authorization access. In order to further improve the accuracy of identity verification, we design an Attention-Enhanced Convolutional Network (AECN) model that integrates the attention mechanism. Specifically, by introducing a confidence score branch, the proposed AECN-based PLA scheme completes authentication without a threshold, thus avoiding the issues stemming from inappropriate threshold settings in conventional PLA schemes. The simulation results show that our proposed AECN framework outperforms existing algorithms at different levels of security authentication.
The Internet's critical infrastructure relies on centralized trust authorities, and their repeated failures expose a structural fragility: these systems lack self-verifiability — the property that any party relying on a claim can independently verify that event or state claim without depending on any third-party authority. Here we propose the Memory-Driven Trust Paradigm, which endows digital infrastructure with self-verifiability through verifiable memory, shifting the fundamental question of trust from "whom to trust" to "what to verify." We define verifiable memory as cryptographically committed, independently verifiable traces of discrete events and states, and formalize it as an "Accumulate–Anchor–Verify" triadic structure: a unified framework neutral to the underlying commitment mechanism, revealing that previously fragmented self-verifiable schemes — blockchains, certificate-transparency logs, and verifiable storage — share the same deep logic. We characterize necessary and sufficient conditions for verifiable memory and give a constructive proof: an epoch hash chain with geometric staged scheduling reduces the worst-case membership witness from Θ(√n) to Θ(n¹ᐟ³) even under unbounded append; we further establish a mechanism-dependence boundary across verification dimensions, closure of constructions under composition, and a verification-baseline reset via anchoring. We instantiate the framework as four kinds of memory — blockchains for transaction memory, message chains for transmission memory, key chains for protection memory, and verifiable AI memory for inter-agent trust — of which the blockchain and its supporting systems have been validated in real-network deployment, and the message chain and key chain have undergone prototype evaluation and comparative experiments. This work establishes self-verifiability as a foundational property of digital infrastructure and provides a verifiable trust foundation for the AI-mediated digital era.