当前位置:首页 > 报告详情

人工智能引发的威胁:了解风险并构建智能防御 [THR2656](2).pdf

上传人: Fl****zo 编号:970959 2025-11-08 22页 1.62MB

1、AI-Induced ThreatsUnderstanding Risks&Building Intelligent Defenses THR2656Vasudev NayakVice President,Technology&Innovation,CLOUDSUFIErin Carlson Senior Principal Product Manager,Oracle1Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted2Copyright 2025,Ora

2、cle and/or its affiliates|Confidential:Internal/Restricted/Highly RestrictedPPT Tips:How to replace images1.Right click image,2.choose CHANGE PICTURE”How to make images have a smaller file size1.Click on the image2.Choose“Picture Format3.Choose“Compress Pictures”Artificial Intelligence is a big step

3、 forward However,weve taken big steps like this before3Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly RestrictedOn-PremisesCloudTechnologyMicroservicesContainersArtificial IntelligenceSuperintelligenceVirtualizationWhat is it that makes AI feel different?THE POWE

4、RAI drives innovation in work,connectivity,and defense.Its a force multiplier for good.FasterExpanded ReachMore SophisticatedTHE PERILAI is an equalizer for bad actors,lowering the barrier to entry for highly sophisticated attacks.New Threat VectorsAI-Powered Threats4Copyright 2025,Oracle and/or its

5、 affiliates|Confidential:Internal/Restricted/Highly RestrictedTHE RESULT:A new threat landscape defined by Speed,Scale,and Sophistication.INDUSTRY PERSPECTIVEHong Kong HeistScenario:Finance worker wired HK$200 million($25.6mm USD)to scammers.Initial Contact-WhatsApp:By individuals impersonating the

6、CFO and a lawyer.Tactics:Previously-scheduled video conference was moved forward on the calendar.Employee targeted with video and audio deepfakes of company personnel,including the CEO.Money was transferred without any indicators of compromise.5Copyright 2025,Oracle and/or its affiliates|Confidentia

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
根据报告的内容,全文主要内容概括如下: 1. **AI带来的威胁与机遇**:AI在推动创新的同时,也降低了攻击门槛,带来了新的威胁向量,如AI驱动的攻击。 2. **案例**:例如,通过深度伪造和身份冒充进行的大额诈骗,以及LLM技术导致的数据泄露事件。 3. **AI安全挑战**:包括模型中毒、提示注入、数据提取、模型倒置等。 4. **安全策略**:采用零信任、分层防御和主动防御策略,如使用Oracle数据库的AI驱动异常检测、Oracle Cloud Guard的机器学习扫描等。 5. **关键点**: - AI安全需要零信任和分层防御。 - 使用原生Oracle安全功能进行控制。 - 通过输入过滤、输出验证和最小权限访问来减少风险。 - 定期审计和人类监督是必要的。 - OWASP AI威胁模型提供了结构化的安全框架。
如何防范?" 揭秘风险与对策" "AI时代,如何构建智能防线?"
客服
商务合作
小程序
服务号
折叠