首页

投稿

大语言模型安全 & LLM漏洞

1307

0

2026-03-12 11:15:00

来源于互联网收集。 == 开启B站官方字幕即可翻译成中文字幕学习， VX订阅 | 哔哩哔哩同名：0x00实验室欢迎关注.

大语言模型

0x00实验室发消息

人无名方能潜心练剑。

1 -Introduction to LLMs and their applications

04:49

2 -Overview of security challenges specific to LLM applications

03:17

3 -Introduction to the OWASP Top 10 LLM Applications list

03:41

4 -Importance of secure LLM development and deployment

04:45

5 -Real-world case studies of successfulunsuccessful LLM implementations

06:13

6 -Common LLM application architectures (e.g., RAG)

06:35

7 -The threat landscape motivations of attackers targeting LLM applications

06:16

1 -Detailed explanation of prompt injection vulnerabilities

03:32

2 -Types of prompt injection (direct and indirect)

04:51

3 -Potential impacts of prompt injection attacks

05:44

4 -Prevention and mitigation strategies

05:54

5 -Evolution of prompt injection techniques and their increasing sophistication

05:28

6 -Impact deep dive specific examples

05:44

7 -Defense-in-depth combining input validation, output filtering, and human revi

05:26

1 -Common examples of vulnerabilities(PII leakage, proprietary algorithm exposur

07:51

2 -Understanding the risks of sensitive information disclosure in LLM applicatio

05:01

3 -Prevention and mitigation strategies (sanitization, access controls, etc.)

06:27

4 -Data minimization importance of minimizing sensitive data collection

05:06

5 -Privacy-enhancing technologies - PET

06:02

6 -Legal and compliance legal implications of sensitive data disclosure

06:56

1 -Supply chain vulnerabilities in LLM development and deployment

05:26

2 -Prevention and mitigation strategies for supply chain risks

04:53

3 -SBOMs in detail explanation of Software Bill of Materials (SBOMs) and their i

05:10

4 -Model provenance challenges difficulties in verifying the origin and integrit

04:51

5 -Governance and policy importance of clear policies for using third-party LLMs

04:28

1 -Understanding data and model poisoning attacks

03:39

2 -How poisoning can impact LLM behavior and security

04:31

3 -Prevention and mitigation strategies

04:48

4 -Poisoning scenarios across the lifecycle poisoning in training and fine-tunin

05:01

5 -Backdoor attacks detail on how backdoors are inserted

04:24

6 -Robustness testing need for rigorous testing to detect poisoning effects

04:54

1 -Risks associated with improper handling of LLM outputs

04:24

2 -Vulnerabilities such as XSS, SQL injection, and remote code execution

04:43

4 -Output encoding examples code examples for different contexts (e.g., HTML, SQ

04:27

5 -Real-world exploits detail cases where improper output handling led to breach

05:19

1 -The concept of agency in LLM systems and associated risks

03:43

2 -Risks of excessive functionality, permissions, and autonomy

04:40

4 -Agentic systems explanation of LLM agents, their benefits, and risks

04:56

5 -Least privilege in depth detailed guidance on implementing least privilege

04:27

6 -Authorization frameworks best practices for managing authorization in LLM

05:10

1 -Vulnerability of system prompt leakage

03:12

2 -Risks associated with exposing system prompts

03:41

4 -Prompt engineering risks how prompt engineering can extract system prompts

05:00

5 -Defense in depth for prompts

05:09

6 -Secure design principles

06:13

1 -Vulnerabilities related to vector and embedding usage in LLM applications

04:34

2 -Risks of unauthorized access, data leakage, and poisoning

05:22

3 -Prevention and mitigation strategies

05:14

4 -Embedding security details on securing vector databases and embeddings

05:35

5 -RAG security best practices

05:47

6 -Emerging research

05:59

1 -The issue of misinformation generated by LLMs

03:50

2 -Causes and potential impacts of misinformation

04:33

3 -Prevention and mitigation strategies

04:58

4 -The spectrum of misinformation

05:32

5 -Impact on specific domains

05:05

6 -Detection and mitigation techniques

05:47

1 -Risks associated with excessive and uncontrolled LLM usage

04:47

2 -Vulnerabilities that can lead to denial of service, economic losses, etc

06:26

3 - Prevention and mitigation strategies

06:02

4 -Economic denial of service

05:09

5 -Rate limiting strategies

05:18

6 -Model extraction defenses

06:14

1 -Summary of key security principles for LLM applications

04:25

2 -Emerging trends and future challenges in LLM security

05:36

3 -Resources and further learning

05:06

4 -Secure LLM development lifecycle integrating security into every stage

05:30

5 -Emerging technologies

06:24

6 -The role of standards and regulations

05:00

顶部