As cities are roused by intelligent sensing capabilities, and each video frame and data point narrates a tale of interconnectedness, a profound inquiry arises: amidst this transformation fueled by the Internet of Things (IoT) and artificial intelligence (AI), how do we preserve trust? How can we guarantee that technology is not only intelligent but also dependable?

On April 18, the 19th China Electronics Information Annual Conference took place in Wuhan. Concurrently, the 2025 China Electronics Institute Science and Technology Awards were officially unveiled.

The project, 'Key Technologies and Applications of AI Security for the Internet of Things,' helmed and primarily accomplished by Hikvision, clinched the first prize of the Science and Technology Progress Award for its remarkable contributions in the realm of IoT AI security. This accolade not only acknowledges the project's technological prowess but also fully endorses Hikvision's IoT security R&D team for their enduring commitment to pioneering technologies and dedication to independent innovation.

Centering on Security Risks: Tackling Challenges in AI and IoT Integration

With the widespread adoption of AI technologies across transportation, security, energy, and other sectors, a multitude of AI models are deployed on edge devices, confronting novel security threats such as data privacy violations, malicious model attacks, and uncontrollable algorithms. Constructing a trustworthy, controllable, and manageable AI system has emerged as a pivotal concern for ensuring the secure functioning of critical infrastructure.

Confronted with industry-wide challenges, Wang Bin, Chief Security Officer of Hikvision, guided the research team to concentrate on the security protection needs of IoT AI. Following years of technological advancements, they erected a comprehensive security protection system across three tiers: trustworthy training, evaluation enhancement, and application control of AI algorithm models, successfully bridging numerous technological gaps both domestically and internationally.

Overcoming Three Major Hurdles: Establishing a Full-Chain Security Protection System

In response to the security challenges posed by AI applications in IoT settings, the project team undertook systematic research at multiple levels.

In the construction of intelligent cities, data pertaining to transportation, security, and the environment held by various departments are frequently isolated, creating 'data silos.' Direct sharing could result in privacy infringements. The project innovatively introduced a highly trustworthy collaborative training and privacy protection approach for multi-domain spatiotemporal data. By ensuring that each participant's data is 'usable but not visible,' it dismantles 'data silos,' facilitating privacy protection and secure sharing of training data in IoT integration scenarios. This guarantees that data yields value while upholding stringent security standards.

If an AI model is infiltrated with malicious samples or subjected to adversarial attacks, it may lead to recognition failures, false positives, or even remote control, with severe repercussions. The team constructed a multi-dimensional security evaluation and protection enhancement system for intelligent IoT terminal models, enabling in-depth detection of algorithm model vulnerabilities and significantly bolstering the model's anti-attack capabilities and robustness while preserving usability.

Numerous edge devices are dispersed throughout cities, with limited resources and decentralized management, making timely anomaly detection challenging. Traditional centralized monitoring approaches often respond sluggishly and fail to meet real-time demands. The team proposed a security risk monitoring and control framework for edge intelligent algorithm applications, substantially enhancing the real-time perception of security threats and precise prevention and control efficiency, ensuring that every edge device operates securely and controllably.

Extensive Application of Outcomes: Bolstering the Intelligent and Secure Evolution of Cities

Presently, the project's outcomes have been extensively implemented across various fields, including public safety, smart energy, and urban management, providing security empowerment for intelligent terminals across diverse industries and constructing a robust security barrier for digital transformation.

As the number of intelligent devices continues to surge and more intelligent scenarios are actualized, security will transition from being an 'optional extra' to a 'necessary component.' Only by advancing technological progress and security protection in tandem can intelligent development be both swift and stable! Hikvision will persist in upholding the philosophy of innovation-driven development, delving deeper into exploration and practice in the cybersecurity domain, and collaborating with partners to jointly construct a secure, trustworthy, and thriving IoT ecosystem!