智能车辆安全数据：重建车辆环境以实现隐私保护机器学习.pdf

1、Smart Vehicles,Secure DataRecreating Vehicle Environments for Privacy-Preserving Machine LearningFrankie CancinoMercedes-Benz Research&Development North AmericaData+AI Summit June 912,2025“It is a fundamental right protected by law.”2European Data Protection Boarddata protection is important.In 2020

2、,California voters approved the CPRA,which added additional privacy protectionsAgenda3Protection&PrivacyPowered by AIPrivacy-Preserving MLSecure by DesignOn the HorizonClosingAgenda4Protection&PrivacyPowered by AIPrivacy-Preserving MLSecure by DesignOn the HorizonClosingSecuring and Governing Data A

3、cross SystemsProtection&PrivacyMercedes-Benz Data Vision6Data handled responsibly and securelySecuring at ScaleIAMNetwork SecurityInfra as Code(IaC)CI/CDMonitoringFrameworks(e.g.GDPR)Data MinimizationFleet-based InsightsNonprod for TestingBest PracticesAnonymizationKey ManagementIn Transit and At Re

4、stAsymmetricEncryption7Integrating Cybersecurity,Data Protection,and PrivacyAgenda8Protection&PrivacyPowered by AIPrivacy-Preserving MLSecure by DesignOn the HorizonClosingSmart Mobility SolutionsPowered by AIRoutines10“New E-Class harnesses AI to learn driver routines”Zero Layer11“Personalized sugg

5、estions for infotainment,comfort and vehicle functions”MBUX Voice Assistant12“Human-like conversations with your Mercedes-Benz”Agenda13Protection&PrivacyPowered by AIPrivacy-Preserving MLSecure by DesignOn the HorizonClosingTechniques and TrendsPrivacy-Preserving MLKey ChallengesThe more you anonymi

6、ze,mask,or restrict access to data,the less useful it becomes for training accurate models.Even if data is anonymized,patterns can still allow someone to re-identify vehicles,especially when combined with other datasets.Model TrainingRe-IdentificationML models can unintentionally memorize sensitive