《轻松实现人工智能安全.pdf》由会员分享,可在线阅读,更多相关《轻松实现人工智能安全.pdf(29页珍藏版)》请在三个皮匠报告上搜索。
1、/SOFTWARE IMPROVEMENT GROUP/SOFTWARE IMPROVEMENT GROUPAI securitymade easyRob van der VeerSANS AI CYBERSECURITY SUMMIT APRIL 1ST2025/SOFTWARE IMPROVEMENT GROUPResponsibleTrustworthyAI writes secure codeFairSafeBiasedDiscriminatingAI verifies securitySkewedGeneralizingTransprarentExplainableSecure AI
2、 systemPrivacy-preservingLawfulEthicalHarmfulAI defendsPerformantRobustFundamental rightsHallucinatingAI attacksStaleAccurate2Security professional,what is your concern?/SOFTWARE IMPROVEMENT GROUPSecurity professional,what is your concern?ResponsibleTrustworthyAI writes secure codeFairSafeBiasedDisc
3、riminatingAI verifies securitySkewedGeneralizingTransprarentExplainableSecure AI systemPrivacy-preservingLawfulEthicalHarmfulAI defendsPerformantRobustFundamental rightsHallucinatingAI attacksStaleAccurate3/SOFTWARE IMPROVEMENT GROUPSecurity professional,what is your concern?ResponsibleTrustworthyAI
4、 writes secure codeFairSafeBiasedDiscriminatingAI verifies securitySkewedGeneralizingTransprarentExplainableSecure AI systemPrivacy-preservingLawfulEthicalHarmfulAI defendsPerformantRobustFundamental rightsHallucinatingAI attacksStaleAccurate4/SOFTWARE IMPROVEMENT GROUPResponsibleTrustworthyAI write
5、s secure codeFairSafeBiasedDiscriminatingAI verifies securitySkewedGeneralizingTransprarentExplainableSecure AI systemPrivacy-preservingLawfulEthicalHarmfulAI defendsPerformantRobustFundamental rightsHallucinatingAI attacksStaleAccurateThe key AI security concern?/SOFTWARE IMPROVEMENT GROUPWhat is s
6、pecial about AI security?6Application&infrastructureInputOutputAIModelAI engineeringStandardengineeringAugmentation dataTraining dataDevelopment-timeRuntime/SOFTWARE IMPROVEMENT GROUPThe six special things about AI securityApplication&infrastructureInputOutputAIModelTraining dataAI engineering7Poiso