Search

In order to study the effect of digital media content production and dissemination supported by artificial intelligence technology, we aim to improve the dissemination and influence of digital media in short video platforms. This paper takes digital media as the research object and launches research on digital media content production and dissemination mode. Based on the research method of inquiry description, the core process of digital media intelligent transformation was deeply analyzed, and the overall logical framework of digital media intelligent transformation was designed. The digital media content production flow model is built by combining data integration and aggregation mining from a digital media content perspective. A graph-convolution matrix decomposition recommendation model for digital media broadcasting was developed by analyzing the performance of algorithmic recommendation in digital media applications. The results are presented: in terms of the order of influence degree size, the degree of influence on the number of likes is greater for political news (Beta = 0.2026, p < 0.01) than for military science and technology (Beta = 0.1637, p < 0.01) than for social news (Beta = 0.1225, p < 0.05), i.e.It indicates that the content theme is shifting from traditional social news to political news, which makes it easier to increase the heat of content dissemination. This study helps us think about the impact of media content production and dissemination strategies in the age of mobile Internet and provides new inspiration for communication theory research.

In this paper, we propose a new DDoS attack detection mechanism based on federated learning that employs dynamic thresholds to cope with the fluctuation of variable rate DDoS attacks. The performance of this detection mechanism is analyzed in terms of traffic classification, performance of verification module, accuracy and loss value. Experimental results show that the method has an accuracy of 99.83% in detecting regular Benign traffic. In burst attack scenarios, the technique significantly improves detection accuracy for all 10 common DDoS attack types. In a sustained attack environment, the intrusion detection system trained based on the DDoS model has the most minor performance degradation, and the average detection accuracy for all types of DDoS attacks still exceeds 90%. Compared with the traditional SVM model, the DDoS attack detection model based on federated learning has a significant performance advantage with Loss and Acc parameters of 0.1 and 0.9, respectively.

This paper explores new approaches to the application and creation of digital media in the context of the rapid development of artificial intelligence technology, with a particular focus on virtual reality interaction. The study utilizes the spatial distance of the enclosing spheres in the hybrid hierarchical enclosing box and the characteristic of non-intersection between the enclosing boxes for collision detection, and the collision information returned by the rays is analyzed in depth. Meanwhile, the intersection test method is used to improve the efficiency of collision detection. And a subjective evaluation of 10 users’ experience in virtual reality art interaction was conducted. The results showed that the system was highly rated in accuracy (average score of 8.39), but rated lower in simplicity (average score of 4.28). There was a positive relationship between the user’s eye gaze duration and the number of interruptions and misoperations in art interactions. At an eye gaze duration of 0.2 seconds, half of the users did not experience interaction interruptions; when the gaze duration increased to 1.2 seconds, all users had no interruptions. This paper provides a practical reference case for using AI technology in digital media applications and creation, especially in enhancing the accuracy and efficiency of virtual reality interactions.