Release Time：2024-11-14

• Share：

• Smart Safeguard is an advanced multimodal large language model (MLLM) and multilingual anti-fraud system launched by ZTE
• ZTE collaborated with Intel to develop a green, power-saving, cost-effective, and data-protecting solution, ensuring the protection of individuals' property and privacy

Shenzhen, China, 14 November 2024 - ZTE Corporation (0763.HK / 000063.SZ), a global leading provider of integrated information and communication technology solutions, has unveiled its Smart Safeguard at the 5G Summit & User Congress 2024 in Istanbul, Türkiye. The system is a green multimodal LLM anti-fraud solution developed in collaboration with Intel, leveraging AI on Xeon CPU.

According to the 2023 annual report from the Global Anti-Scam Alliance (GASA), online fraud resulted in a loss of approximately USD 1.02 trillion globally, which equates to 1.05% of global GDP, a significant increase from USD 55.3 billion in 2022. Telecom network services, including SMS, MMS, and voice/video calls, are widely used across the globe, with around 9.3 billion users, making them prime targets for internet fraud.

Fraud is evolving rapidly, with tactics changing frequently, making it harder to detect using traditional static rules-based systems. Fraudulent telecom networks are overwhelming the system, leading to substantial economic and social consequences. As fraud schemes evolve, the need for dynamic and intelligent fraud detection has never been greater.

Introducing ZTE Smart Safeguard
ZTE Smart Safeguard is the first new-generation anti-fraud system based on multimodal large language models (MLLM). It integrates large language models, computer vision models, and multimodal technology to effectively detect and process a variety of content types, including text, images, audio, and video. This technology supports multiple languages and dialects, including English, French, Spanish, Chinese, Burmese, Cantonese, and Fujian dialects.

The system has been proven to be highly intelligent, efficient, accurate, and adaptable, offering continuous self-upgrades to combat evolving fraud techniques.

Collaborating with Intel for Green Computing
Large language models (LLMs) require significant computing power, which traditionally involves expensive, high-power GPUs. These can present challenges such as high costs, difficult procurement, and excessive energy consumption. In response to these challenges, ZTE and Intel have jointly developed the first green MLLM anti-fraud solution powered by AI on Xeon CPUs.

Key benefits of the solution include:

·AI computing power integrated into CPUs: Intel's Xeon Scalable processors feature built-in AI accelerators (Intel AMX), which improve the performance of deep learning tasks such as NLP and image recognition.

·Green emission reduction and energy savings: The solution eliminates the need for large GPUs and additional cooling infrastructure, saving space and energy.

·Low total cost of ownership (TCO): The solution reduces hardware and operational costs, offering improved ROI through efficient use of CPU resources.

·AI security protection: Intel's SGX technology ensures application-level encryption and data protection within a secure enclave.

·Simple deployment and maintenance: The solution's easy-to-deploy design requires no complicated network configurations, reducing maintenance efforts and ensuring system stability.

Enhanced Security, Robustness, and Interoperability
Smart Safeguard offers comprehensive security features, ensuring privacy and data protection. It has high robustness, with redundant architecture and fine-tuning capabilities to ensure high service availability and accuracy. The system is also designed for easy integration with any carrier network, offering scalability and broad application potential.

Proven Results
Through the joint laboratory and real-world network data, the solution has demonstrated impressive results:

·Text accuracy and recall rates: 99%

·Multimedia content accuracy: Over 95%, surpassing the performance of traditional systems

·Power consumption reduction: 40% savings in energy use for AI services

·Space savings: Reduction of servers and racks by 30%

·Cost savings: 45% reduction in TCO for hybrid deployment

Smart Safeguard has already been deployed in many provincial branches of China Mobile, China Unicom, and other operators. Compared to traditional systems, it has reduced fraud cases by 64%, improved accuracy by 44%, and decreased error rates by 36%. The system also saves more than 80% in manual audit costs, boosting operational efficiency and driving economic benefits.

Social Value and Future Impact
Beyond its technical capabilities, Smart Safeguard enhances user experience by providing secure, reliable communication services. It contributes to public property protection, social stability, and overall customer satisfaction.

As a pioneering green anti-fraud solution, ZTE's collaboration with Intel introduces a new direction for sustainable AI governance. Moving forward, ZTE plans to expand the application of this innovative solution to more regions globally, making a significant contribution to the green, secure, and sustainable development of network communication.