国家级高新技术企业 National High-Tech Enterprise
国际/国家标准的引领者 Leader in Int'l & National Standards

定义智能感知的安全边界 Defining the Security Boundary of Intelligent Perception

面向智能汽车与机器人的对抗攻击检测、防御及隐私计算解决方案 Adversarial attack detection, defense, and privacy computing solutions for autonomous vehicles and robotics.

探索解决方案Solutions 预约演示Book a Demo

Technology

感知对抗攻击与隐私匿名化 Perception Adversarial Attack & Privacy Anonymization

依托清华海峡研究院联合实验室，我们将学术界最前沿的攻防理论及技术转化为智能汽车、机器人等领域的安全屏障。 Leveraging our Joint Lab with Tsinghua Strait Research Institute, we translate cutting-edge attack-defense theories into security barriers for smart cars and robots.

Adversarial Attack and Defense Technology
ATTACK & DEFENSE

多模态感知攻击及防御算法 Multi-modal Perception Attack & Defense

基于多尺度特征抑制的2D-3D联合对抗攻击策略，自主研发对抗攻击算法、样本生成平台及防御评测基准，在攻击成功率、迁移性、多角度领域全球领先。 Based on multi-scale feature suppression for 2D-3D joint attacks. We develop proprietary algorithms and benchmarks, leading globally in attack success rates, transferability, and multi-angle domains.

Privacy Anonymization Comparison
PRIVACY AI

深度特征保留匿名化技术 Deep Feature Retention Anonymization (DFRA)

DFRA是基于生成式 AI 的隐私保护技术，在实现身份不可逆向还原的同时，精准保留年龄、性别、行为意图等核心分析特征，完全符合 EuroPriSe 合规标准。 DFRA is a generative AI-based privacy tech. While ensuring irreversible identity protection, it accurately retains core analytical features (age, gender, intent), fully compliant with EuroPriSe standards.

Products

全栈式安全解决方案 Full-Stack Security Solutions

Inspection Device

感知系统对抗攻击检测设备及平台 Perception Adversarial Attack Detection Equipment & Platform

第三方测评设备或部署终端的安全哨兵。它能够通过黑盒/白盒的方式针对智能汽车或机器人的感知系统开展对抗攻击，支持C-NCAP/ICAP/GCAP以及L2强标。 Third-party testing equipment for terminal security. Supports Black-box/White-box attacks on perception systems, compatible with C-NCAP/ICAP/GCAP and L2 mandatory standards.

  • 兼容主流检测模型 (YOLO, PointPillars, Sparse4D)Compatible with YOLO, PointPillars, Sparse4D
  • 支持不同类型的激光雷达（905nm, 940nm, 1550nm）Supports LiDAR (905nm, 940nm, 1550nm)
  • 支持多模态及单模态（如纯视觉、激光雷达、语音等）对抗攻击Supports multimodal (vision, LiDAR, voice) attacks
Simulation Engine

感知系统对抗攻击训练及仿真平台 Perception Adversarial Attack Training & Simulation

专为自动驾驶设计的全栈式对抗仿真与脆弱性测评平台。通过高精度仿真引擎，对各种对抗场景下的感知鲁棒性进行深度量化评估。 Full-stack adversarial simulation and vulnerability assessment platform for autonomous driving. Deeply evaluate perception robustness in complex scenarios.

  • 全场景高精度仿真与对抗样本生成（集成Carla）High-fidelity simulation & sample generation (Carla)
  • 多维度智能测评引擎与攻击防御加固Intelligent evaluation & defense hardening engine
  • 深度适配 NuScenes 等主流增强现实公开数据集Deep adaptation to NuScenes and mainstream datasets
Compliance

图像视频匿名化与检测系统 Video Anonymization Platform

满足 GDPR 与 PIPL 合规工具。在保证自动驾驶训练数据可用性的前提下，实现深度特征保留的图像视频匿名化。 Compliance tool for GDPR & PIPL. Ensures data utility for CV training while providing deep feature retention anonymization.

  • 支持 4K 高分辨率视频流实时处理Supports real-time 4K video stream processing
  • ISO 27001 / 27701 标准深度合规ISO 27001 / 27701 deep compliance

Innovation Lab

"洞悉"感知对抗攻击实验室 "Insight" Perception Adversarial Attack Lab

万觉科技始终站在行业前沿，致力于构建开放的技术生态。我们联合清华大学、北京航空航天大学及多家领军企业，共同打造了全国首个感知对抗实验室-"洞悉"。 WanJue Technology always stands at the forefront of the industry, committed to building an open technical ecosystem. In collaboration with Tsinghua University, Beihang University, and multiple leading enterprises, we have co-founded the nation's first perception adversarial lab - "Insight".

实验室专注于感知系统对抗样本的深度研究与实战演练，具备涵盖单/多模态攻击测试、红蓝对抗攻防演习及系统风险挖掘等核心能力，旨在为智能感知安全筑起坚实屏障。 The lab focuses on deep research and practical drills of adversarial samples for perception systems, possessing core capabilities in single/multi-modal attack testing, red-blue adversarial drills, and risk mining.

欢迎预约参观交流。 Welcome to book a visit and exchange.

Testing
攻击测试Attack Testing
Drills
红蓝对抗Red-Blue Drills
Mining
风险挖掘Risk Mining
Training
样本训练Sample Training

关于万觉科技 About WanJue.Ai

万觉科技成立于2020年，是一家专注于智能网联汽车与具身智能机器人"感知安全"的国家级高新技术企业。公司由清华大学创新领军工程博士团队创立，两院院士担任首席指导专家。 Founded in 2020, WanJue Technology is a National High-Tech Enterprise focusing on the "Perception Security" of Intelligent Connected Vehicles (ICV) and Embodied AI Robotics. The company was founded by an Innovation Leading Engineering PhD team from Tsinghua University, with academicians from the Chinese Academy of Sciences and Engineering serving as chief advisors.

作为感知安全领域的领航者，万觉科技牵头国家首个智能网联汽车领域的感知安全标准，并荣获2025年"先导杯"智能车辆领域一等奖。我们拥有20余项自主发明专利，并构建了全国首家"洞悉"感知对抗攻击实验室。目前，公司已通过ISO 9001、ISO 45001、ISO 14001等管理体系认证，持续为多家主流车企及国家级检测机构提供卓越的安全保障服务。 As a pioneer in perception security, WanJue Technology spearheaded the nation's first perception security standard for ICVs and was awarded 1st Prize in the 2025 "Xiandao Cup" in the intelligent vehicle category. We hold over 20 proprietary invention patents and have established China's first "Insight" Perception Adversarial Attack Lab. Currently, the company is ISO 9001/45001/14001 certified, providing exceptional security services to numerous major automotive OEMs and national-level testing institutions.

1st
先导杯一等奖Xiandao Cup 1st Prize
20+
发明专利Patents
3
ISO 管理体系ISO Systems

资质与荣誉 Certifications & Honors

2025"先导杯"一等奖2025 Xiandao Cup 1st Prize
国际/国家标准牵头单位Standard Setting Leader
ISO 9001 / 27001 / 45001

开始构建更安全的感知系统 Build Safer Perception Systems

无论您是整车厂、Tier 1 供应商还是研究机构，我们都随时准备为您提供支持。 Whether you are an OEM, Tier 1 supplier, or research institution, we are here to help.

sales@wanjueai.com
中国 · 北京Beijing, China