GenAI能否解决多云安全问题？.pdf

1、Can GenAI Solve Multicloud Security?Featuring material from SEC510:Cloud Security Engineering and Controls sans.org/sec510Can GenAI Solve Multicloud Security?GenAI(In)Effective Security Use CasesMulticloudComplexityThe ExperimentAttempts and ResultsSecurity RisksLive DemoAdditional ResourcesQ&AAgend

2、aGenAI(In)EffectiveSecurity Use CasesCan GenAI Solve Multicloud Security?Relevant LLM Terminology GenAI:Artificial intelligence that can create data not found in its training data.LLMs:Models applying GenAI to Natural Language Processing(NLP).Traditional AI:Types of AI commonly applied prior to the

3、rise of GenAI.Prompt:A question or request provided to the GenAI model.Hallucinations:Incorrect responses from GenAI platforms.The AI will often appear confident in its wrong answers,even if they are nonsensical.Temperature:Parameter affecting the randomness of a GenAIs response.Higher temperatures

4、mean more creative responses.Can GenAI Solve Multicloud Security?Applying LLMs to Security Ineffective Use Cases Organizations are quick to apply new,hyped-up technologies to every problem.Some security use cases for GenAI are more promising than others.Well-defined problems that require little or n

5、o context are often better solved with traditional AI than LLMs.Chess bots have been much better than humans for decades,while LLMs are much weaker in comparison and constantly hallucinate illegal moves.Software Composition Analysis(SCA)uses a repeatable process without much human interaction.It als

6、o depends on real-time data,while LLM datasets are stale.Static/Dynamic Application Security Testing(SAST/DAST)processes are more complicated,but they are the same regardless of the organization using them.Using LLMs to search through documentation can provide less accurate results than traditional