为应对 AI 深度伪造带来的安全风险，蚂蚁数科联合高校设计了一种全新的学习框架，可以将图像信息 中的纹理、颜色、边缘等特征解耦并完成多尺度特征分析，大大提升深度伪造的"解题"难度。与此同 时，蚂蚁数科提供一种"关键遗忘"机制，可以让检测模型忽略常见的伪造痕迹，转而去捕捉更深层的伪 造特征，从而提升对未知伪造手法的识别与预警能力。 蚂蚁数科与华东师范大学合作，在零知识证明领域取得两项关键突破，提出构建了业界首个可验证图数 据查询算法和业界首个可验证联邦学习框架ZKSL，为企业间跨组织的智能体协作构建一个可信的底层 协议层，以快速生成验证证明，确保协同决策的可信与高效。 记者注意到，蚂蚁数科研究成果覆盖NeurIPS、USENIX Security、CCS、NDSS、ICDE等多个前沿领域 的国际顶级会议与期刊，标志着相关技术成果达到所属领域的全球领先水平。与此同时，蚂蚁数科与清 华大学、浙江大学、华东师范大学等高校建立了联合实验室、定向科研课题等深度合作机制，通过产学 研协同持续推动前沿技术探索与落地。 晓风 在 AI 时代，企业在迈向智能化的道路上正面临着三大关键挑战：如何在分布式环境中建立信任、如何 应对A ...

为破解城市交通拥堵痛点，神思电子紧扣"数字济南"战略与数智警务体系建设要求，聚焦主干道与快速 路拥堵治理，构建"多维感知—动态建模—自适应控制"的数智治理新范式。 通过整合多源大数据，优化交通感知网与信号控制体系，实现"拥堵自动感知、事故自动发现、信号智 能优化、诱导精准服务"，助力济南打造全国领先的智能信控示范体系和现代化数智交管标杆，切实提 升市民出行体验。 近日，神思电子技术股份有限公司（以下简称"神思电子"）在济南数字化建设领域捷报频传，中标领域 覆盖城市安全、数据要素、智慧交通、低空经济领域。公司以硬核技术实力深度参与"数字济南"战略落 地，成为城市数字化转型的核心合作伙伴。 深耕城市安全领域构建数智交管新生态 近年来，神思电子(300479)在"一体两翼"战略的引领下，依托深厚的行业资源积淀与领先的人工智能技 术优势，将AI技术赋能千行百业，其中城市智慧更新领域是公司业务着力点。 公司通过全链条技术赋能城市治理，依托物联网感知层建设、异构数据融合分析及AI决策中枢构建， 实现从基础设施智能化到城市治理精细化的跃升。 在城市安全与应急救援建设领域，神思电子携手浪潮智慧城市科技有限公司，基于济南市现有 ...

2024年5月，平台在海尔、美的、小天才等10余家典型代表企业开展应用试点，在针对智能摄像头的实 际应用场景检测中，发现了高、中危害漏洞12项，及时提出了基于可信计算的轻量化端云联动信息安全 防御的解决方法，大幅降低了该类产品的安全风险隐患。 未来，团队将进一步适配智能机器人、低空飞行器等新兴产品，构建"靶向检测-智能预警-精准处置"全 链条数智化服务平台，推动我国智能产品质量安全从"被动应对"向"主动防控"转型，夯实产业安全发展 根基。（中国经济网记者 佟明彪） 据悉，研究团队依托5000多条历史检测数据和200多项标准训练细分领域垂直大模型，成功构建自动生 成测试用例并自动执行的智能体，内置10余项标准检测模块，覆盖100余项检测项目，兼容10种通信协 议，可支持12类智能产品"即插即测"，效率较传统人工检测提升20倍，并支持检测报告自动化审核。 智能家居行业信息安全检测领域临着传统工具覆盖指标少、测试效率低、恶意攻击辨识难等痛点。对 此，市场监管总局组织研究团队开发了"智能家居信息安全异构融合检测平台"。该平台融合搭建了质量 安全领域大模型（质言大模型），为智能家电、智能安防、智能穿戴等产品提供全方位、 ...

Core Viewpoint - The 2025 Autonomous Trusted Computing Innovation Forum was held in Beijing, highlighting the importance of trusted computing and cybersecurity in China's digital economy, with Anheng Information becoming a key player in the collaborative innovation mechanism for autonomous trusted computing [1][4]. Industry Overview - The construction of the cybersecurity grading protection system is a significant initiative in China's cybersecurity field, with trusted computing serving as a core technological support for building proactive defense systems and ensuring the controllability of information systems [3][5]. Company Highlights - Anheng Information has extensive experience in the cybersecurity field, contributing to grading protection construction, trusted computing technology research and development, product implementation, and application in various scenarios [3][6]. - The company recently launched the "Grading Protection Evaluation Report Analysis Intelligent Agent," which integrates AI and core standards to create a multi-dimensional intelligent review engine, marking a new era in intelligent report auditing [6]. - Anheng Information was the first company to complete the initial "Enhanced Trusted Execution Environment Basic Capability Special Test" organized by the China Academy of Information and Communications Technology, focusing on privacy computing technologies [6]. - The company aims to build a secure data flow infrastructure and develop trusted data space products, utilizing advanced technologies such as Big Data Trusted Execution Environment (BDTEE), Multi-Party Computation (MPC), and federated learning to ensure reliable, controllable, and traceable multi-party data joint computing processes [6].

Core Viewpoint - The company is developing a trusted AI inference card based on its affiliate Jiangyuan Technology's domestically produced AI computing power chips and its own CCT02 TCM trusted chips, which supports national cryptography algorithms and enhances security against potential threats [1] Group 1 - The trusted AI inference card supports national cryptography algorithms SM2/3/4 [1] - It features static boot measurement capabilities to prevent software-related attacks and protect sensitive information [1] - The product complies with the GM/T 0012-2020 trusted computing standard [1]

Core Insights - The collaboration between Zhongke Shuguang and China Mobile aims to address the inefficiencies in data storage and computing power allocation in the "East Data West Computing" initiative [1][2] - The current investment ratio in computing power versus storage in China is approximately 40:1, compared to around 10:1 in foreign markets, indicating a significant imbalance [1][2] - Shuguang Storage has developed a unified scheduling platform to enhance data flow and management across different data centers, addressing key challenges in data visibility, organization, and utilization [3][4][5] Group 1: Industry Challenges - The primary bottleneck in AI infrastructure is not computing power but rather the foundational data layer, which is often fragmented and incompatible across different systems [3][4] - Data centers face issues such as lack of visibility into data assets, inefficient data flow, and high latency in accessing stored data, particularly in the context of the "East Data West Computing" strategy [4][5] - The rapid growth of AI applications has increased the performance requirements for data storage systems, necessitating advancements in IOPS and bandwidth capabilities [6][12] Group 2: Technological Advancements - Shuguang Storage has transitioned from being a follower to a solution provider in the storage technology space, leveraging over 21 years of R&D to establish a competitive edge [2][10] - The company has achieved global recognition for its centralized storage products, ranking first in the SPC-1 international benchmark for storage performance [2][10] - The development of a unified storage scheduling platform is seen as essential for facilitating data flow across heterogeneous and geographically dispersed data centers [7][8] Group 3: Strategic Collaborations - The partnership with China Mobile was initiated to address specific business needs and to support the national strategy of optimizing resource utilization in the West [7][8] - Shuguang Storage's extensive self-research capabilities position it as a key player in solving complex data flow challenges, which many other vendors struggle to address [8][9] Group 4: Future Outlook - The entry barriers for AI applications are expected to lower with advancements in technology, but new challenges related to privacy and data security will emerge [12][13] - Enhancing GPU resource utilization is critical for reducing AI operational costs, with Shuguang Storage's innovations in data transfer efficiency playing a pivotal role [13][14] - The company's focus on "storage-computing synergy" aims to optimize the interaction between storage systems and GPU resources, significantly improving processing speeds and efficiency [14]