当前位置:首页 > 报告详情

高级人工智能安全:为人工智能工作负载构建纵深防御体系.pdf

上传人: 明**** 编号:1013072 2025-12-21 64页 659.03KB

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.S E C 4 1 0Advanced AI Security:Architecting Defense-in-Depth for AI WorkloadsRiggs Goodman IIIPrincipal SA AI SecurityAWSJason GarmanPrincipal Security SAAWS 2025,A

2、mazon Web Services,Inc.or its affiliates.All rights reserved.AgendaPhase 1:Foundational ModelsPhase 2:Data SourcesPhase 3:ToolsPhase 4:AgentsConclusionThought bubbles:Important,thought-provoking questions will be here.QR Codes 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.This i

3、s a L400 talk,and we will show code forAmazon Bedrock AgentCoreManaged AWS service that provides infrastructure and tools to securely deploy and operate AI agents at scaleAgentsAmazon BedrockFully managed AWS service that offers a choice of high-performing foundation models(FMs)Generative AIStrandsA

4、 model-driven approach to building AI agents in just a few lines of codeAgent FrameworkWhy code?Code makes it explainable vs a black box 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.B E D R O C K A G E N T C O R EAgentCore ServicesAmazon Bedrock AgentCore enables you to deploy

5、and operate highly effective agents securely,at scale using any framework and modelAgentCore RuntimeA secure,serverless runtime purpose-built for deploying and scaling dynamic AI agents and toolsregardless of framework,protocol,or model choice.AgentCore IdentityProvides a secure,scalable agent ident

6、ity and access management capability accelerating AI agent development.AgentCore GatewayProvides a secure way for agents to discover and use tools along with easy transformation of APIs,Lambda functions,and existing services into agent-compatible tools.AgentCore MemoryMakes it easy for developers to

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. **AWS Bedrock AgentCore**:一个全托管服务,提供基础设施和工具,以安全、大规模地部署和操作AI代理。 2. **基础模型**:AWS Bedrock提供多种高性能基础模型(FMs)供选择。 3. **数据来源**:数据可来自外部,如工具API、知识库和用户输入。 4. **工具**:工具允许AI应用和代理执行外部操作,如调用API或控制浏览器。 5. **代理**:AI代理是自主软件程序,可与环境交互,执行自我指导的任务。 6. **安全措施**:包括验证输入/输出、身份验证、授权和实施安全策略。 7. **安全风险**:关注未授权数据、外部访问、敏感数据和高风险操作。
如何构建防御深度?" AI安全部署新利器?" 自主行动背后的安全挑战?"
客服
商务合作
小程序
服务号
折叠