Collective perception and swarm intelligence collaborative learning have emerged as innovative paradigms in sensing and computation, pivotal for enhancing the efficacy of the industrial internet. However, these advancements have also brought to light a plethora of cybersecurity issues within industrial control systems, with the leakage of industrial data and user privacy information becoming a prominent concern. Traditional security measures are ill-equipped to address the novel security and privacy risks posed by the extensive interconnectivity of the future industrial internet. Consequently, this work introduces a high-performance collaborative data analysis chain with privacy protection. It achieves this through the optimization of distributed deep learning performance for consortium blockchains and the implementation of efficient swarm intelligence perception based on Bayesian differential privacy, thereby ensuring the privacy protection and high-performance analysis of data within the industrial internet.