MindSpore Becomes World's First Trustworthy AI Framework Certified with CC EAL2+
MindSpore Becomes World's First Trustworthy AI Framework Certified with CC EAL2+
Recently, the all-scenario AI framework MindSpore has officially passed the security assessment of the world's largest independent security evaluation lab — SGS Brightsight, obtaining the Common Criteria Evaluation Assurance Level 2+ certified configuration (CC EAL2+) certification. As the first CC certification received by an AI framework worldwide, it marks that MindSpore has made great achievements and won global recognition in information security.
Nowadays, the world is witnessing tremendous changes in technology. AI technologies, featuring deep learning, cross-industry convergence, human-machine collaboration, and independent control, are driving a new round of technological revolution and industrial transformation. While making our lives easier and more comfortable, these technologies also pose great public concern due to technical features such as generality, algorithm black box, and data dependency. As a result, specialized, diversified, agile, and globalized AI governance is in urgent need. Against this backdrop, trustworthy AI comes into being and has become a major research field in the international community.
Regulators and researchers have widely discussed the connotation and extension of trustworthy AI. Based on the discussions, we have concluded the following trustworthy technologies throughout the AI lifecycle:
MindSpore, based on the AI lifecycle, has built the following technical capabilities.
1. Model security: Protects model with robustness evaluation, adversarial evaluation and training, and model encryption.
2. Privacy protection: Develops technologies such as privacy assessment, differential privacy training, and federated learning to ensure data privacy.
3. Explainability: Improves the explainability of models with the innovative semantically interpretable technologies and toolsets for users to better understand, trust, and use models.
CC certification is currently one of the most authoritative product safety certifications in the world. The evaluation standard is based on IEC/ISO15408 Common Criteria for Information Technology Security Evaluation, which is the general criteria for information technology security evaluation. MindSpore, this time, is evaluated by SGS Brightsight, an authoritative product information security test lab designated by the Netherlands.
During the evaluation, MindSpore worked with SGS Brightsight to innovatively set the security objectives of AI frameworks, which provide standards for future evaluation programs. This is made possible because of MindSpore's long-time research in trustworthy AI.
Mr. Zhang Kaifan, Chief Operating Officer of SGS Brightsight in Asia, said: "As the world's leading security certification lab, SGS Brightsight is more than happy to work with MindSpore to write the history of security certification. AI is playing a more and more important part in our daily lives. However, the definition and implementation of AI products and their security functions have not been determined. Through this program, MindSpore and SGS Brightsight have proved that certification for AI frameworks is possible and will be necessary in the future."
In the era of intelligent transformation, MindSpore is taking trustworthy AI to a new height, enhancing AI security in terms of security technologies and security governance. In the future, MindSpore will keep innovating and work together with partners to address AI security challenges.
MindSpore is a revolutionary AI framework for device, edge, and cloud scenarios. It aims to build a new AI programming paradigm, allowing you to develop better, efficient, and flexible AI software and hardware applications. In addition to version iteration and feature optimization, MindSpore is dedicated to building an interactive learning community.
Join us to unleash the power of AI with MindSpore.
Official website: https://mindspore.cn/en