Traditional rule-based intrusion detection methods struggle to address evolving threats, such as zero-day attacks, while deep learning models remain vulnerable to adversarial perturbations. Moreover, flattening structured data may result in the loss of spatial and temporal information. To address these limitations, this paper proposes a spatiotemporal detection model that integrates a graph convolutional network, a long short-term memory network, multi-head attention fusion, and min-max adversarial training based on projected gradient descent. Experimental results show that the proposed model achieves an F1 score of 97.42% on the clean test set and maintains F1 scores of 88.64% and 81.33% under white-box projected gradient descent attacks with ε = 0.1 and ε = 0.2, respectively. The proposed model outperforms the convolutional neural network and long short-term memory baselines.
With the popularization of intelligent recommendation on online news platforms, centralized data collection leads to privacy risks, and the requirements for algorithm compliance are increasingly strict. In view of the difficulties in implementing privacy protection scheme, balancing effect, and security, and separating technology from governance, the authors of this paper constructed a multi-cycle data flow and privacy risk model, quantitatively evaluated the whole link in stages, and designed a training service strategy and a security audit state machine that integrates federal learning and differential privacy. Experiments showed that this scheme could reduce the privacy risk index to 0.42, control the click-through rate and ranking quality loss within 1.5%, slightly improve the coverage rate of hot news, and significantly enhance users' privacy awareness, platform trust, and willingness to use, thus providing a feasible path for news platforms to carry out personalized recommendation safely and in compliance.
This study focuses on the dynamic adaptation problem of hierarchical protection of personal information in complex businesses. It breaks through the limitations of traditional linear processes to build a dynamic mechanism for personal information protection that integrates multi-loop feedback and state machines. The study supports the ternary matrix of information type‒protection level‒risk attribute; it also supports the whole process risk quantification system. Relying on closed-loop architecture, two-way feedback, and node redundancy to achieve dynamic adjustment, the study introduces a comprehensive scoring model dominated by probability feedback, risk update, and highly sensitive data. The empirical results show that this mechanism can reduce the system failure rate and the compliance rate increase, the Level 4 violation detection rate is increased to 0.39, and the alarm efficiency is increased.
Digitalization of intangible cultural heritage (ICH) intellectual property (IP) suffers from ambiguous ownership and difficulties in effective brand asset accumulation of digital media art works. This study applies blockchain to ICH IP rights confirmation, develops a traceable brand asset evaluation mechanism for digital media art works, and builds a multistakeholder collaborative confirmation model. It clarifies the roles of on-chain evidence, smart contracts, and reputation feedback mechanisms; establishes an evaluation system combining on-chain mechanism indicators and brand asset variables; and empirically examines the relationship between blockchain-based rights confirmation and ICH brand performance through correlation, regression, and scenario analyses. Finally, a blockchain-driven ICH IP rights confirmation and brand asset evaluation model is proposed, offering insights for ICH digital rights management and sustainable brand development.
Self-sovereign-identity systems grant users control over their digital identities, thereby eliminating the need for centralized authorities. Conventional blockchain-based identity management systems can expose identity-related metadata while imposing significant storage and computational overhead, and this raises issues with maintaining privacy and scalability. A hybrid architecture of “blockchain + zero-knowledge proof” was proposed by this study for privacy-first self-sovereign-identity verification, where the layers of public and private blockchains were integrated with zero-knowledge proofs to assure a secure, scalable, and privacy-preserving identity verification while protecting sensitive information from exposure. The proposed system had less data exposure, 0.2% against 10% in traditional systems. It also had a higher verification speed that could reach a one identity verification in 0.0051 s at low loads against 0.0068 s in traditional settings. Hence, the proposed system might double the performance, 98 transactions processed per second against 49 transactions processed per second, with improved efficiency hovering between 20% and 40%.
