Unclecheng-li/DeepSec

http://youtube.com/watch?v=7ZA0Z1R-MjQ

http://youtube.com/watch?v=KksYizcLFH0

[Credential-related source omitted]

https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks

https://atlas.mitre.org/techniques/AML.T0051

https://github.com/invariantlabs-ai/mcp-injection-experiments

https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks

https://mp.weixin.qq.com/s/EJLb1IwqbPF3VSDkJu099g

https://x.com/hongming731/status/1922261630664245326

https://news.qq.com/rain/a/20250429A07QY000

https://blog.trailofbits.com/2025/04/21/jumping-the-line-how-mcp-servers-can-attack-you-before-you-ever-use-them/

https://blog.trailofbits.com/2025/04/29/deceiving-users-with-ansi-terminal-codes-in-mcp/

https://blog.trailofbits.com/2025/04/29/deceiving-users-with-ansi-terminal-codes-in-mcp/

https://www.solo.io/blog/deep-dive-mcp-and-a2a-attack-vectors-for-ai-agents

https://aclanthology.org/2024.scalellm-1.2/

https://atlas.mitre.org/techniques/AML.T0051

https://josephthacker.com/ai/2023/05/19/prompt-injection-poc.html

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/SSRF.html

https://systemweakness.com/new-prompt-injection-attack-on-chatgpt-web-version-ef717492c5c2

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Unauthorized_Code_Execution.html

https://www.calvin-risk.com/blog/decoding-llm-risks-a-comprehensive-look-at-unauthorized-code-execution

https://mp.weixin.qq.com/s/eFDQWYYCOe_SSiourhTxig

https://arxiv.org/html/2402.16914v1

https://promptengineering.org/mind-over-malware-battling-the-growing-arsenal-of-attacks-on-large-language-models/

https://www.toolify.ai/ai-news/the-future-of-hacking-5-terrifying-llm-security-threats-544868

https://embracethered.com/blog/posts/2024/chatgpt-persistent-denial-of-service/

https://openai.com/index/memory-and-new-controls-for-chatgpt/

https://portswigger.net/web-security/llm-attacks

https://labs.withsecure.com/publications/llm-agent-prompt-injection

https://labs.withsecure.com/publications/llm-agent-prompt-injection

http://youtube.com/watch?v=7ZA0Z1R-MjQ

http://youtube.com/watch?v=KksYizcLFH0

https://mp.weixin.qq.com/s/9JwADiu9t3kqcfqnRMC2zQ

https://finance.sina.com.cn/tech/digi/2025-06-01/doc-ineypqvh0855918.shtml

https://zhuanlan.zhihu.com/p/1900540531131523166

https://mp.weixin.qq.com/s/2bm7nuXkORLZ20mfpOmwrA

https://atlas.mitre.org/techniques/AML.T0051.001

https://twitter.com/random_walker/status/1636923058370891778

https://medium.com/@harry.hphu/introduction-to-web-llm-attacks-indirect-prompt-injection-7bb9f154bc07

https://medium.com/@dinob5551/indirect-prompt-injection-the-hidden-threat-lurking-in-ai-730b009dd5fb

https://drive.google.com/file/d/1CTVcliUblX35cWfB49Xjhf8xk-fM3QH1/edit?pli=1

https://analyticsindiamag.com/metas-llama-leaked-to-the-public-thanks-to-4chan/

https://knightcolumbia.org/blog/the-llama-is-out-of-the-bag-should-we-expect-a-tidal-wave-of-disinformation

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://cloud.baidu.com/article/3253170

https://www.akto.io/blog/insecure-output-handling-in-llms-insights

https://journal.hexmos.com/insecure-output-handling/

https://systemweakness.com/new-prompt-injection-attack-on-chatgpt-web-version-ef717492c5c2

https://sec.cafe/handbook/security_research/ai_security/llm_security/attack/

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/SSRF.html

https://www.horizon3.ai/attack-research/attack-blogs/nextchat-an-ai-chatbot-that-lets-you-talk-to-anyone-you-want-to/

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://www.wehelpwin.com/article/5063

https://medium.com/nfactor-technologies/rag-poisoning-an-emerging-threat-in-ai-systems-660f9ff279f9

https://ironcorelabs.com/security-risks-rag/

https://arxiv.org/html/2312.04724v1

https://security.tencent.com/index.php/blog/msg/209

https://hiddenlayer.com/research/insane-in-the-supply-chain/

https://atlas.mitre.org/techniques/AML.T0029

https://owasp.org/www-project-top-10-for-large-language-model-applications/assets/PDF/OWASP-Top-10-for-LLMs-2023-v05.pdf

https://www.cnblogs.com/LittleHann/p/17596696.html

https://blog.securelayer7.net/owasp-top10-for-large-language-models/

https://www.mufeedvh.com/llm-security/#2-sandboxing-extended-llms

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_Sandboxing.html

https://mp.weixin.qq.com/s/tf4ljSJ0Ue0YniojWhYMKg

https://www.wiz.io/blog/wiz-and-hugging-face-address-risks-to-ai-infrastructure

https://pradiptabanerjee.medium.com/confidential-containers-for-large-language-models-42477436345a

https://www.run.ai/guides/kubernetes-architecture/securing-your-ai-ml-kubernetes-environment

https://mp.weixin.qq.com/s/Ry1PoZLfPvw6Lj8bz14mgw

https://github.com/knownsec/KCon/blob/master/2023/CICD%E6%94%BB%E5%87%BB%E5%9C%BA%E6%99%AF.pdf

https://xie.infoq.cn/article/536a3e7e776eb32b38d1a9747

https://www.helloaliyun.com/tutorial/1039.html

https://support.huaweicloud.com/usermanual-gaussdbformysql/gaussdbformysql_05_0347.html

https://developer.aliyun.com/article/1430094

https://pradiptabanerjee.medium.com/confidential-containers-for-large-language-models-42477436345a

https://ironcorelabs.com/security-risks-rag/

https://www.securityweek.com/chatgpt-data-breach-confirmed-as-security-firm-warns-of-vulnerable-component-exploitation/

http://www.bimant.com/blog/top8-ml-model-deployment-tools/

https://mlflow.org/docs/latest/deployment/index.html

https://www.docker.com/blog/llm-docker-for-local-and-hugging-face-hosting/

https://collabnix.com/large-language-models-llms-and-docker-building-the-next-generation-web-application/

https://mp.weixin.qq.com/s/vIDHBLbA5iWoPlYTKHSZfw

https://cloud.baidu.com/article/621826

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_Sandboxing.html

https://www.secrss.com/articles/64006

https://huntr.com/bounties/a795bf93-c91e-4c79-aae8-f7d8bda92e2a

https://doc.dataiku.com/dss/latest/concepts/homepage/index.html

https://www.secrss.com/articles/62742

https://llmtop10.com/llm05/

- https://cloud.baidu.com/article/621826

[Credential-related source omitted]

| 目标 | 检测命令 | 利用方式 | |------|----------|----------| | 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 |

|------|----------|----------| | 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 | | K8s API | `curl -k https://kubernetes.default.svc` | 列举Pod/获取Secret |

| 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 |

| K8s API | `curl -k https://kubernetes.default.svc` | 列举Pod/获取Secret |

| K8s ServiceAccount | `cat /var/run/secrets/kubernetes.io/serviceaccount/token` | 认证K8s API | | 内网数据库 | `echo \| nc DB_HOST 5432` | 直连数据库 | | Redis | `redis-cli -h REDIS_HOST ping` | 未授权访问 |

| 内网数据库 | `echo \| nc DB_HOST 5432` | 直连数据库 |

| Docker Registry | `curl http://REGISTRY:5000/v2/_catalog` | 拉取敏感镜像 |

https://nordicapis.com/llm-security-hinges-on-api-security/

https://superface.ai/blog/how-to-connect-openai-gpts-to-apis

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://www.secrss.com/articles/51763

https://github.com/linexjlin/GPTs

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://twitter.com/simonw/status/1570933190289924096

https://magazine.sebastianraschka.com/p/ahead-of-ai-8-the-latest-open-source

https://vulcan.io/blog/owasp-top-10-llm-risks-what-we-learned/#h2_1

https://www.linkedin.com/pulse/security-threats-around-llm-systems-categorization-gaurang-desai-bvale?trk=article-ssr-frontend-pulse_more-articles_related-content-card

https://www.anquanke.com/post/id/247895

https://www.aixinzhijie.com/article/6825834

https://blog.barracuda.com/2024/04/03/generative-ai-data-poisoning-manipulation

https://36kr.com/p/2723023103489920

https://shardsecure.com/blog/data-manipulation-ml

https://blog.csdn.net/2401_84252820/article/details/138406655?utm_medium=distribute.pc_relevant.none-task-blog-2~default~baidujs_baidulandingword~default-4-138406655-blog-124579765.235v43pc_blog_bottom_relevance_base5&spm=1001.2101.3001.424

https://cloud.baidu.com/article/3248650

https://forum.butian.net/share/3072

https://aws.amazon.com/cn/blogs/china/privacy-and-security-of-agent-applications/

https://arxiv.org/abs/2511.17671?utm_source=chatgpt.com

https://arxiv.org/abs/2601.05504?utm_source=chatgpt.com

https://www.lesswrong.com/posts/aPeJE8bSo6rAFoLqg/solidgoldmagikarp-plus-prompt-generation

https://www.lesswrong.com/posts/Ya9LzwEbfaAMY8ABo/solidgoldmagikarp-ii-technical-details-and-more-recent

https://www.nightfall.ai/ai-security-101/model-inversion

https://www.michalsons.com/blog/model-inversion-attacks-a-new-ai-security-risk/64427

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://bj.bcebos.com/ensec-web-privacy/anquan/%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%AE%89%E5%85%A8%E8%A7%A3%E5%86%B3%E6%96%B9%E6%A1%88%E7%99%BD%E7%9A%AE%E4%B9%A6.pdf

https://mp.weixin.qq.com/s/JlJwDRzYG985kF4d6g7qjw

https://news.cctv.com/2022/06/21/ARTIdhgLL1sSK5Hjl0uYWybr220621.shtml

https://atlas.mitre.org/techniques/AML.T0036

https://www.kuaikuaicloud.com/market/3667.html

http://www.nelab-bdst.org.cn/data/upload/ueditor/20230707/64a78209c719c.pdf

https://mp.weixin.qq.com/s/3WAWy4ZV6Ezft_2MJHMgtg

https://mp.weixin.qq.com/s/yiloJtlmv7MT3df9AnWNZQ

https://mp.weixin.qq.com/s/c_cIzecyw48MatwKBZbdUg

https://36kr.com/p/2541963790493187

https://mp.weixin.qq.com/s/VCmhL-LbGfCViQrAEwyCAg

https://mp.weixin.qq.com/s/kp1Sl5TC_uuVelhj8HPmdw

https://mp.weixin.qq.com/s/VCmhL-LbGfCViQrAEwyCAg

https://mp.weixin.qq.com/s/kp1Sl5TC_uuVelhj8HPmdw

https://mp.weixin.qq.com/s/c_cIzecyw48MatwKBZbdUg

https://36kr.com/p/2541963790493187

https://platform.openai.com/docs/guides/fine-tuning/preparing-your-dataset

https://arxiv.org/abs/2310.03693

https://blog.csdn.net/yalecaltech/article/details/117135011

https://cloud.baidu.com/article/1819998

https://www.eet-china.com/mp/a213535.html

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Training_Data_Poisoning.html

https://mp.weixin.qq.com/s/C9eIW06UXKL8g9TkZzGn_w

https://www.techpolicy.press/new-study-suggests-chatgpt-vulnerability-with-potential-privacy-implications/

https://ensarseker1.medium.com/data-poisoning-attacks-the-silent-threat-to-ai-integrity-d83900eea276

https://www.51cto.com/article/760084.html

https://home.dartmouth.edu/news/2024/01/zeroing-origins-bias-large-language-models

https://mp.weixin.qq.com/s/lgMI9tf0xAl8siZYaKcqog

https://mcp.csdn.net/6800a595a5baf817cf49422d.html

https://www.reddit.com/r/cybersecurity/comments/1lzrkf6/another_critical_cvss_9610_mcpbased_vulnerability/

https://threatprotect.qualys.com/2025/07/03/anthropic-model-context-protocol-mcp-inspector-remote-code-execution-vulnerability-cve-2025-49596/?utm_source=chatgpt.com

https://arxiv.org/pdf/2211.09527.pdf

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://prompt-guide.xiniushu.com/prompt_hacking/injection

https://www.knightcxx.cn/?p=118

https://simonwillison.net/2023/Feb/15/bing/

https://www.tomshardware.com/news/chatgpt-generates-windows-11-pro-keys

https://www.polygon.com/23690187/discord-ai-chatbot-clyde-grandma-exploit-chatgpt?continueFlag=9d7655502c6eb54decc775fab724139d

https://sysdig.com/blog/lateral-movement-cloud-containers/

https://dtzed.com/studies/2023/10/8093/

https://www.cobalt.io/blog/llm-insecure-output-handling

https://allabouttesting.org/owasp-agentic-ai-threat-t9-identity-spoofing-impersonation-in-ai-systems/

https://moanju.org/posts/ai-agent-attack-examples-owasp-2026/

https://openai.com/blog/march-20-chatgpt-outage

https://securityaffairs.com/144057/data-breach/openai-chatgpt-redis-bug-data-leak.html

https://kenhuangus.medium.com/llm-powered-applications-architecture-patterns-and-security-controls-7a153c3ec9f4

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Insufficient_Access_Control.html

https://mp.weixin.qq.com/s/DMx-By1qxB5cQglkaq9ppQ

https://priyalwalpita.medium.com/securing-the-future-of-ai-a-deep-dive-into-owasps-top-10-security-risks-for-large-language-models-72c5ff540cd3

http://www.nelab-bdst.org.cn/data/upload/ueditor/20230707/64a78209c719c.pdf

https://blog.csdn.net/douyu0814/article/details/133703803

https://www.knightcxx.cn/?p=118

https://thehackernews.com/2023/06/over-100000-stolen-chatgpt-account.html

https://www.makeuseof.com/why-hackers-target-chatgpt-accounts/

https://mp.weixin.qq.com/s/DMx-By1qxB5cQglkaq9ppQ

https://www.signalfire.com/blog/prompt-injection-security

https://developer.nvidia.com/blog/mitigating-stored-prompt-injection-attacks-against-llm-applications/

- https://www.securityweek.com/major-organizations-using-hugging-face-ai-tools-put-at-risk-by-leaked-api-tokens/

- https://aws.amazon.com/cn/what-is/api-key/

https://medium.com/@nitishjoshi060291/llm-hallucinations-fix-it-with-vector-database-de04eee531da

https://cloudsecurityalliance.org/blog/2023/11/22/mitigating-security-risks-in-retrieval-augmented-generation-rag-llm-applications

https://www.cnblogs.com/LittleHann/p/17440063.html#_label3

https://dongnian.icu/llms/llms_article/9.%E6%A3%80%E7%B4%A2%E5%A2%9E%E5%BC%BALLM/index.html

https://cloudsecurityalliance.org/blog/2023/11/22/mitigating-security-risks-in-retrieval-augmented-generation-rag-llm-applications

https://www.leewayhertz.com/security-in-ai-development/

https://atmosphericthinking.medium.com/massive-leak-of-chatgpt-credentials-over-100-000-affected-db6cef3a18c5

https://blog.csdn.net/FreeBuf_/article/details/140870185?utm_relevant_index=7

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://developer.nvidia.com/zh-cn/blog/securing-llm-systems-against-prompt-injection/

https://blog.csdn.net/qq_43543209/article/details/135683986

https://www.pulumi.com/ai/answers/mptvxaHguJ6A4yXSHi92zZ/implementing-role-based-access-to-ai-training-data-in-snowflake

https://github.com/0xk1h0/ChatGPT_DAN

https://www.digitaltrends.com/computing/what-is-dan-prompt-chatgpt/

https://arxiv.org/abs/2308.03825

https://www.anthropic.com/research/many-shot-jailbreaking

https://www.lakera.ai/blog/guide-to-hallucinations-in-large-language-models

https://arxiv.org/pdf/2305.13534.pdf

https://atlas.mitre.org/techniques/AML.T0005

https://mp.weixin.qq.com/s/LSTZUKOlXP9VZTxa-nKkhA

https://blog.uptrain.ai/llm-jailbreak/

https://www.fuzzylabs.ai/blog-post/jailbreak-attacks-on-large-language-models

https://www.lakera.ai/blog/jailbreaking-large-language-models-guide

https://mp.weixin.qq.com/s/EhEqNlIcpu9RZ36XFL3vWQ

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://arxiv.org/abs/2307.15043

https://twitter.com/andyzou_jiaming/status/1684766170766004224

https://zhuanlan.zhihu.com/p/662098517

https://zhuanlan.zhihu.com/p/620575831

https://atlas.mitre.org/techniques/AML.T0015

https://mp.weixin.qq.com/s/yozvoCG92TDIF86EEz9g8Q

https://mp.weixin.qq.com/s/RdIQBaBR0RQJUFp0Pf7ovA

https://mp.weixin.qq.com/s/sxjU930eO4K_HKPPWXPlWg

https://mp.weixin.qq.com/s/PGMVqjeI18x7GZyksvtGzQ

https://arxiv.org/pdf/2311.05232.pdf

https://mp.weixin.qq.com/s/qFAQQJ_FuhY2iaLzkoWynA

https://www.lakera.ai/blog/guide-to-hallucinations-in-large-language-models

https://www.appendata.com/blogs/ai-hallucinations

https://mp.weixin.qq.com/s/4UzoMtIL2oSkxzzuceuxhg

https://zh-cn.eureporter.co/internet-2/artificial-intelligence/2024/02/03/laws-to-prevent-ai-terrorism-are-urgently-needed/

https://infosecwriteups.com/jail-breaking-chatgpt-to-write-malware-9b3ae111f30c

https://www.theregister.com/2024/04/17/gpt4_can_exploit_real_vulnerabilities/

https://arxiv.org/abs/2404.08144

https://blog.csdn.net/pengpengjy/article/details/132478358

https://www.freebuf.com/articles/ai-security/454527.html

https://zhuanlan.zhihu.com/p/1928583554805260699

https://mp.weixin.qq.com/s/5cEkxtEbH7GUKiQ5aRsnrg

https://mp.weixin.qq.com/s/nOn1aQDEQys5D7sNK1_oPg

https://mp.weixin.qq.com/s/ZpM09SUHSTvM9SrvrlBEmA

https://www.ibm.com/topics/model-drift

https://www.datacamp.com/tutorial/understanding-data-drift-model-drift

https://mp.weixin.qq.com/s/QbADBoHEqpDBKNkr-so3Ig

https://arxiv.org/pdf/2307.09009.pdf

https://arxiv.org/abs/2404.12038

https://github.com/PKU-YuanGroup/Hallucination-Attack

https://zhuanlan.zhihu.com/p/661444210

https://arxiv.org/pdf/2310.01469.pdf

https://mp.weixin.qq.com/s/EhEqNlIcpu9RZ36XFL3vWQ

http://www.cbdio.com/BigData/2024-01/11/content_6176237.htm

https://mp.weixin.qq.com/s/krG4NGBCjK5kUL2x_RGRSg

https://world.huanqiu.com/article/4FIHGhZ96FB

https://insight.kellogg.northwestern.edu/zh/article/how-to-prepare-for-ai-generated-misinformation

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://mp.weixin.qq.com/s/KGqu6i2_xX9d7-x8P189Lw

https://mp.weixin.qq.com/s/8Ca4HmkafP9SxjHayC9zdQ

https://mp.weixin.qq.com/s/-0i0SlGat-Y5hXcM3EIGiw

https://mp.weixin.qq.com/s/2Ai4nKOzEnkhqJD903O8mA

https://mp.weixin.qq.com/s/2bm7nuXkORLZ20mfpOmwrA

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://36kr.com/p/2190993024614530

https://atlas.mitre.org/tactics/AML.TA0001

https://www.sohu.com/a/584853485_121124363

https://atlas.mitre.org/techniques/AML.T0014

https://atlas.mitre.org/techniques/AML.T0013

https://36kr.com/p/2653630408081670

https://www.sciencedirect.com/science/article/abs/pii/S0167865522003063

https://rodtrent.substack.com/p/must-learn-ai-security-part-8-model

https://arxiv.org/pdf/2403.06634.pdf

https://cloud.tencent.com/developer/article/2378846

https://www.53ai.com/news/LargeLanguageModel/2024071740891.html

https://atlas.mitre.org/techniques/AML.T0018

https://defence.ai/ai-security/backdoor-attacks-ml/

https://arxiv.org/abs/2308.14367

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_AI_Alignment.html

https://wiki.offsecml.com/Supply+Chain+Attacks/Models/Using+Keras+Lambda+Layers

https://5stars217.github.io/2023-08-08-red-teaming-with-ml-models/

https://splint.gitbook.io/cyberblog/security-research/tensorflow-remote-code-execution-with-malicious-model

https://thenewstack.io/how-to-reduce-the-hallucinations-from-large-language-models/

https://aclanthology.org/2020.acl-main.249/

http://youtube.com/watch?v=7ZA0Z1R-MjQ

http://youtube.com/watch?v=KksYizcLFH0

[Credential-related source omitted]

https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks

https://atlas.mitre.org/techniques/AML.T0051

https://github.com/invariantlabs-ai/mcp-injection-experiments

https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks

https://mp.weixin.qq.com/s/EJLb1IwqbPF3VSDkJu099g

https://x.com/hongming731/status/1922261630664245326

https://news.qq.com/rain/a/20250429A07QY000

https://blog.trailofbits.com/2025/04/21/jumping-the-line-how-mcp-servers-can-attack-you-before-you-ever-use-them/

https://blog.trailofbits.com/2025/04/29/deceiving-users-with-ansi-terminal-codes-in-mcp/

https://blog.trailofbits.com/2025/04/29/deceiving-users-with-ansi-terminal-codes-in-mcp/

https://www.solo.io/blog/deep-dive-mcp-and-a2a-attack-vectors-for-ai-agents

https://aclanthology.org/2024.scalellm-1.2/

https://atlas.mitre.org/techniques/AML.T0051

https://josephthacker.com/ai/2023/05/19/prompt-injection-poc.html

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/SSRF.html

https://systemweakness.com/new-prompt-injection-attack-on-chatgpt-web-version-ef717492c5c2

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Unauthorized_Code_Execution.html

https://www.calvin-risk.com/blog/decoding-llm-risks-a-comprehensive-look-at-unauthorized-code-execution

https://mp.weixin.qq.com/s/eFDQWYYCOe_SSiourhTxig

https://arxiv.org/html/2402.16914v1

https://promptengineering.org/mind-over-malware-battling-the-growing-arsenal-of-attacks-on-large-language-models/

https://www.toolify.ai/ai-news/the-future-of-hacking-5-terrifying-llm-security-threats-544868

https://embracethered.com/blog/posts/2024/chatgpt-persistent-denial-of-service/

https://openai.com/index/memory-and-new-controls-for-chatgpt/

https://portswigger.net/web-security/llm-attacks

https://labs.withsecure.com/publications/llm-agent-prompt-injection

https://labs.withsecure.com/publications/llm-agent-prompt-injection

http://youtube.com/watch?v=7ZA0Z1R-MjQ

http://youtube.com/watch?v=KksYizcLFH0

https://mp.weixin.qq.com/s/9JwADiu9t3kqcfqnRMC2zQ

https://finance.sina.com.cn/tech/digi/2025-06-01/doc-ineypqvh0855918.shtml

https://zhuanlan.zhihu.com/p/1900540531131523166

https://mp.weixin.qq.com/s/2bm7nuXkORLZ20mfpOmwrA

https://atlas.mitre.org/techniques/AML.T0051.001

https://twitter.com/random_walker/status/1636923058370891778

https://medium.com/@harry.hphu/introduction-to-web-llm-attacks-indirect-prompt-injection-7bb9f154bc07

https://medium.com/@dinob5551/indirect-prompt-injection-the-hidden-threat-lurking-in-ai-730b009dd5fb

https://drive.google.com/file/d/1CTVcliUblX35cWfB49Xjhf8xk-fM3QH1/edit?pli=1

https://analyticsindiamag.com/metas-llama-leaked-to-the-public-thanks-to-4chan/

https://knightcolumbia.org/blog/the-llama-is-out-of-the-bag-should-we-expect-a-tidal-wave-of-disinformation

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://cloud.baidu.com/article/3253170

https://www.akto.io/blog/insecure-output-handling-in-llms-insights

https://journal.hexmos.com/insecure-output-handling/

https://systemweakness.com/new-prompt-injection-attack-on-chatgpt-web-version-ef717492c5c2

https://sec.cafe/handbook/security_research/ai_security/llm_security/attack/

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/SSRF.html

https://www.horizon3.ai/attack-research/attack-blogs/nextchat-an-ai-chatbot-that-lets-you-talk-to-anyone-you-want-to/

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://www.wehelpwin.com/article/5063

https://medium.com/nfactor-technologies/rag-poisoning-an-emerging-threat-in-ai-systems-660f9ff279f9

https://ironcorelabs.com/security-risks-rag/

https://arxiv.org/html/2312.04724v1

https://security.tencent.com/index.php/blog/msg/209

https://hiddenlayer.com/research/insane-in-the-supply-chain/

https://atlas.mitre.org/techniques/AML.T0029

https://owasp.org/www-project-top-10-for-large-language-model-applications/assets/PDF/OWASP-Top-10-for-LLMs-2023-v05.pdf

https://www.cnblogs.com/LittleHann/p/17596696.html

https://blog.securelayer7.net/owasp-top10-for-large-language-models/

https://www.mufeedvh.com/llm-security/#2-sandboxing-extended-llms

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_Sandboxing.html

https://mp.weixin.qq.com/s/tf4ljSJ0Ue0YniojWhYMKg

https://www.wiz.io/blog/wiz-and-hugging-face-address-risks-to-ai-infrastructure

https://pradiptabanerjee.medium.com/confidential-containers-for-large-language-models-42477436345a

https://www.run.ai/guides/kubernetes-architecture/securing-your-ai-ml-kubernetes-environment

https://mp.weixin.qq.com/s/Ry1PoZLfPvw6Lj8bz14mgw

https://github.com/knownsec/KCon/blob/master/2023/CICD%E6%94%BB%E5%87%BB%E5%9C%BA%E6%99%AF.pdf

https://xie.infoq.cn/article/536a3e7e776eb32b38d1a9747

https://www.helloaliyun.com/tutorial/1039.html

https://support.huaweicloud.com/usermanual-gaussdbformysql/gaussdbformysql_05_0347.html

https://developer.aliyun.com/article/1430094

https://pradiptabanerjee.medium.com/confidential-containers-for-large-language-models-42477436345a

https://ironcorelabs.com/security-risks-rag/

https://www.securityweek.com/chatgpt-data-breach-confirmed-as-security-firm-warns-of-vulnerable-component-exploitation/

http://www.bimant.com/blog/top8-ml-model-deployment-tools/

https://mlflow.org/docs/latest/deployment/index.html

https://www.docker.com/blog/llm-docker-for-local-and-hugging-face-hosting/

https://collabnix.com/large-language-models-llms-and-docker-building-the-next-generation-web-application/

https://mp.weixin.qq.com/s/vIDHBLbA5iWoPlYTKHSZfw

https://cloud.baidu.com/article/621826

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_Sandboxing.html

https://www.secrss.com/articles/64006

https://huntr.com/bounties/a795bf93-c91e-4c79-aae8-f7d8bda92e2a

https://doc.dataiku.com/dss/latest/concepts/homepage/index.html

https://www.secrss.com/articles/62742

https://llmtop10.com/llm05/

- https://cloud.baidu.com/article/621826

[Credential-related source omitted]

| 目标 | 检测命令 | 利用方式 | |------|----------|----------| | 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 |

|------|----------|----------| | 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 | | K8s API | `curl -k https://kubernetes.default.svc` | 列举Pod/获取Secret |

| 云元数据 | `curl 169.254.169.254` | 获取IAM临时凭据 |

| K8s API | `curl -k https://kubernetes.default.svc` | 列举Pod/获取Secret |

| K8s ServiceAccount | `cat /var/run/secrets/kubernetes.io/serviceaccount/token` | 认证K8s API | | 内网数据库 | `echo \| nc DB_HOST 5432` | 直连数据库 | | Redis | `redis-cli -h REDIS_HOST ping` | 未授权访问 |

| 内网数据库 | `echo \| nc DB_HOST 5432` | 直连数据库 |

| Docker Registry | `curl http://REGISTRY:5000/v2/_catalog` | 拉取敏感镜像 |

https://nordicapis.com/llm-security-hinges-on-api-security/

https://superface.ai/blog/how-to-connect-openai-gpts-to-apis

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://www.secrss.com/articles/51763

https://github.com/linexjlin/GPTs

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://learnprompting.org/docs/prompt_hacking/leaking

https://simonwillison.net/2022/Sep/12/prompt-injection/

https://matt-rickard.com/a-list-of-leaked-system-prompts

https://genai.stackexchange.com/questions/197/how-to-effectively-prevent-prompt-leaking-via-injection

https://twitter.com/simonw/status/1570933190289924096

https://magazine.sebastianraschka.com/p/ahead-of-ai-8-the-latest-open-source

https://vulcan.io/blog/owasp-top-10-llm-risks-what-we-learned/#h2_1

https://www.linkedin.com/pulse/security-threats-around-llm-systems-categorization-gaurang-desai-bvale?trk=article-ssr-frontend-pulse_more-articles_related-content-card

https://www.anquanke.com/post/id/247895

https://www.aixinzhijie.com/article/6825834

https://blog.barracuda.com/2024/04/03/generative-ai-data-poisoning-manipulation

https://36kr.com/p/2723023103489920

https://shardsecure.com/blog/data-manipulation-ml

https://blog.csdn.net/2401_84252820/article/details/138406655?utm_medium=distribute.pc_relevant.none-task-blog-2~default~baidujs_baidulandingword~default-4-138406655-blog-124579765.235v43pc_blog_bottom_relevance_base5&spm=1001.2101.3001.424

https://cloud.baidu.com/article/3248650

https://forum.butian.net/share/3072

https://aws.amazon.com/cn/blogs/china/privacy-and-security-of-agent-applications/

https://arxiv.org/abs/2511.17671?utm_source=chatgpt.com

https://arxiv.org/abs/2601.05504?utm_source=chatgpt.com

https://www.lesswrong.com/posts/aPeJE8bSo6rAFoLqg/solidgoldmagikarp-plus-prompt-generation

https://www.lesswrong.com/posts/Ya9LzwEbfaAMY8ABo/solidgoldmagikarp-ii-technical-details-and-more-recent

https://www.nightfall.ai/ai-security-101/model-inversion

https://www.michalsons.com/blog/model-inversion-attacks-a-new-ai-security-risk/64427

https://mp.weixin.qq.com/s/ygqRv4vGW5YZS1SiVzAejg

https://bj.bcebos.com/ensec-web-privacy/anquan/%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%AE%89%E5%85%A8%E8%A7%A3%E5%86%B3%E6%96%B9%E6%A1%88%E7%99%BD%E7%9A%AE%E4%B9%A6.pdf

https://mp.weixin.qq.com/s/JlJwDRzYG985kF4d6g7qjw

https://news.cctv.com/2022/06/21/ARTIdhgLL1sSK5Hjl0uYWybr220621.shtml

https://atlas.mitre.org/techniques/AML.T0036

https://www.kuaikuaicloud.com/market/3667.html

http://www.nelab-bdst.org.cn/data/upload/ueditor/20230707/64a78209c719c.pdf

https://mp.weixin.qq.com/s/3WAWy4ZV6Ezft_2MJHMgtg

https://mp.weixin.qq.com/s/yiloJtlmv7MT3df9AnWNZQ

https://mp.weixin.qq.com/s/c_cIzecyw48MatwKBZbdUg

https://36kr.com/p/2541963790493187

https://mp.weixin.qq.com/s/VCmhL-LbGfCViQrAEwyCAg

https://mp.weixin.qq.com/s/kp1Sl5TC_uuVelhj8HPmdw

https://mp.weixin.qq.com/s/VCmhL-LbGfCViQrAEwyCAg

https://mp.weixin.qq.com/s/kp1Sl5TC_uuVelhj8HPmdw

https://mp.weixin.qq.com/s/c_cIzecyw48MatwKBZbdUg

https://36kr.com/p/2541963790493187

https://platform.openai.com/docs/guides/fine-tuning/preparing-your-dataset

https://arxiv.org/abs/2310.03693

https://blog.csdn.net/yalecaltech/article/details/117135011

https://cloud.baidu.com/article/1819998

https://www.eet-china.com/mp/a213535.html

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Training_Data_Poisoning.html

https://mp.weixin.qq.com/s/C9eIW06UXKL8g9TkZzGn_w

https://www.techpolicy.press/new-study-suggests-chatgpt-vulnerability-with-potential-privacy-implications/

https://ensarseker1.medium.com/data-poisoning-attacks-the-silent-threat-to-ai-integrity-d83900eea276

https://www.51cto.com/article/760084.html

https://home.dartmouth.edu/news/2024/01/zeroing-origins-bias-large-language-models

https://mp.weixin.qq.com/s/lgMI9tf0xAl8siZYaKcqog

https://mcp.csdn.net/6800a595a5baf817cf49422d.html

https://www.reddit.com/r/cybersecurity/comments/1lzrkf6/another_critical_cvss_9610_mcpbased_vulnerability/

https://threatprotect.qualys.com/2025/07/03/anthropic-model-context-protocol-mcp-inspector-remote-code-execution-vulnerability-cve-2025-49596/?utm_source=chatgpt.com

https://arxiv.org/pdf/2211.09527.pdf

https://www.packtpub.com/article-hub/preventing-prompt-attacks-on-llms

https://prompt-guide.xiniushu.com/prompt_hacking/injection

https://www.knightcxx.cn/?p=118

https://simonwillison.net/2023/Feb/15/bing/

https://www.tomshardware.com/news/chatgpt-generates-windows-11-pro-keys

https://www.polygon.com/23690187/discord-ai-chatbot-clyde-grandma-exploit-chatgpt?continueFlag=9d7655502c6eb54decc775fab724139d

https://sysdig.com/blog/lateral-movement-cloud-containers/

https://dtzed.com/studies/2023/10/8093/

https://www.cobalt.io/blog/llm-insecure-output-handling

https://allabouttesting.org/owasp-agentic-ai-threat-t9-identity-spoofing-impersonation-in-ai-systems/

https://moanju.org/posts/ai-agent-attack-examples-owasp-2026/

https://openai.com/blog/march-20-chatgpt-outage

https://securityaffairs.com/144057/data-breach/openai-chatgpt-redis-bug-data-leak.html

https://kenhuangus.medium.com/llm-powered-applications-architecture-patterns-and-security-controls-7a153c3ec9f4

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Insufficient_Access_Control.html

https://mp.weixin.qq.com/s/DMx-By1qxB5cQglkaq9ppQ

https://priyalwalpita.medium.com/securing-the-future-of-ai-a-deep-dive-into-owasps-top-10-security-risks-for-large-language-models-72c5ff540cd3

http://www.nelab-bdst.org.cn/data/upload/ueditor/20230707/64a78209c719c.pdf

https://blog.csdn.net/douyu0814/article/details/133703803

https://www.knightcxx.cn/?p=118

https://thehackernews.com/2023/06/over-100000-stolen-chatgpt-account.html

https://www.makeuseof.com/why-hackers-target-chatgpt-accounts/

https://mp.weixin.qq.com/s/DMx-By1qxB5cQglkaq9ppQ

https://www.signalfire.com/blog/prompt-injection-security

https://developer.nvidia.com/blog/mitigating-stored-prompt-injection-attacks-against-llm-applications/

- https://www.securityweek.com/major-organizations-using-hugging-face-ai-tools-put-at-risk-by-leaked-api-tokens/

- https://aws.amazon.com/cn/what-is/api-key/

https://medium.com/@nitishjoshi060291/llm-hallucinations-fix-it-with-vector-database-de04eee531da

https://cloudsecurityalliance.org/blog/2023/11/22/mitigating-security-risks-in-retrieval-augmented-generation-rag-llm-applications

https://www.cnblogs.com/LittleHann/p/17440063.html#_label3

https://dongnian.icu/llms/llms_article/9.%E6%A3%80%E7%B4%A2%E5%A2%9E%E5%BC%BALLM/index.html

https://cloudsecurityalliance.org/blog/2023/11/22/mitigating-security-risks-in-retrieval-augmented-generation-rag-llm-applications

https://www.leewayhertz.com/security-in-ai-development/

https://atmosphericthinking.medium.com/massive-leak-of-chatgpt-credentials-over-100-000-affected-db6cef3a18c5

https://blog.csdn.net/FreeBuf_/article/details/140870185?utm_relevant_index=7

https://genai.owasp.org/wp-content/uploads/2024/05/OWASP-Top-10-for-LLM-Applications-v1_1_Chinese.pdf

https://developer.nvidia.com/zh-cn/blog/securing-llm-systems-against-prompt-injection/

https://blog.csdn.net/qq_43543209/article/details/135683986

https://www.pulumi.com/ai/answers/mptvxaHguJ6A4yXSHi92zZ/implementing-role-based-access-to-ai-training-data-in-snowflake

https://github.com/0xk1h0/ChatGPT_DAN

https://www.digitaltrends.com/computing/what-is-dan-prompt-chatgpt/

https://arxiv.org/abs/2308.03825

https://www.anthropic.com/research/many-shot-jailbreaking

https://www.lakera.ai/blog/guide-to-hallucinations-in-large-language-models

https://arxiv.org/pdf/2305.13534.pdf

https://atlas.mitre.org/techniques/AML.T0005

https://mp.weixin.qq.com/s/LSTZUKOlXP9VZTxa-nKkhA

https://blog.uptrain.ai/llm-jailbreak/

https://www.fuzzylabs.ai/blog-post/jailbreak-attacks-on-large-language-models

https://www.lakera.ai/blog/jailbreaking-large-language-models-guide

https://mp.weixin.qq.com/s/EhEqNlIcpu9RZ36XFL3vWQ

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://arxiv.org/abs/2307.15043

https://twitter.com/andyzou_jiaming/status/1684766170766004224

https://zhuanlan.zhihu.com/p/662098517

https://zhuanlan.zhihu.com/p/620575831

https://atlas.mitre.org/techniques/AML.T0015

https://mp.weixin.qq.com/s/yozvoCG92TDIF86EEz9g8Q

https://mp.weixin.qq.com/s/RdIQBaBR0RQJUFp0Pf7ovA

https://mp.weixin.qq.com/s/sxjU930eO4K_HKPPWXPlWg

https://mp.weixin.qq.com/s/PGMVqjeI18x7GZyksvtGzQ

https://arxiv.org/pdf/2311.05232.pdf

https://mp.weixin.qq.com/s/qFAQQJ_FuhY2iaLzkoWynA

https://www.lakera.ai/blog/guide-to-hallucinations-in-large-language-models

https://www.appendata.com/blogs/ai-hallucinations

https://mp.weixin.qq.com/s/4UzoMtIL2oSkxzzuceuxhg

https://zh-cn.eureporter.co/internet-2/artificial-intelligence/2024/02/03/laws-to-prevent-ai-terrorism-are-urgently-needed/

https://infosecwriteups.com/jail-breaking-chatgpt-to-write-malware-9b3ae111f30c

https://www.theregister.com/2024/04/17/gpt4_can_exploit_real_vulnerabilities/

https://arxiv.org/abs/2404.08144

https://blog.csdn.net/pengpengjy/article/details/132478358

https://www.freebuf.com/articles/ai-security/454527.html

https://zhuanlan.zhihu.com/p/1928583554805260699

https://mp.weixin.qq.com/s/5cEkxtEbH7GUKiQ5aRsnrg

https://mp.weixin.qq.com/s/nOn1aQDEQys5D7sNK1_oPg

https://mp.weixin.qq.com/s/ZpM09SUHSTvM9SrvrlBEmA

https://www.ibm.com/topics/model-drift

https://www.datacamp.com/tutorial/understanding-data-drift-model-drift

https://mp.weixin.qq.com/s/QbADBoHEqpDBKNkr-so3Ig

https://arxiv.org/pdf/2307.09009.pdf

https://arxiv.org/abs/2404.12038

https://github.com/PKU-YuanGroup/Hallucination-Attack

https://zhuanlan.zhihu.com/p/661444210

https://arxiv.org/pdf/2310.01469.pdf

https://mp.weixin.qq.com/s/EhEqNlIcpu9RZ36XFL3vWQ

http://www.cbdio.com/BigData/2024-01/11/content_6176237.htm

https://mp.weixin.qq.com/s/krG4NGBCjK5kUL2x_RGRSg

https://world.huanqiu.com/article/4FIHGhZ96FB

https://insight.kellogg.northwestern.edu/zh/article/how-to-prepare-for-ai-generated-misinformation

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://mp.weixin.qq.com/s/KGqu6i2_xX9d7-x8P189Lw

https://mp.weixin.qq.com/s/8Ca4HmkafP9SxjHayC9zdQ

https://mp.weixin.qq.com/s/-0i0SlGat-Y5hXcM3EIGiw

https://mp.weixin.qq.com/s/2Ai4nKOzEnkhqJD903O8mA

https://mp.weixin.qq.com/s/2bm7nuXkORLZ20mfpOmwrA

https://stcn.com/article/detail/1250289.html

https://www.51cto.com/aigc/912.html

https://36kr.com/p/2190993024614530

https://atlas.mitre.org/tactics/AML.TA0001

https://www.sohu.com/a/584853485_121124363

https://atlas.mitre.org/techniques/AML.T0014

https://atlas.mitre.org/techniques/AML.T0013

https://36kr.com/p/2653630408081670

https://www.sciencedirect.com/science/article/abs/pii/S0167865522003063

https://rodtrent.substack.com/p/must-learn-ai-security-part-8-model

https://arxiv.org/pdf/2403.06634.pdf

https://cloud.tencent.com/developer/article/2378846

https://www.53ai.com/news/LargeLanguageModel/2024071740891.html

https://atlas.mitre.org/techniques/AML.T0018

https://defence.ai/ai-security/backdoor-attacks-ml/

https://arxiv.org/abs/2308.14367

https://owasp.org/www-project-top-10-for-large-language-model-applications/Archive/0_1_vulns/Inadequate_AI_Alignment.html

https://wiki.offsecml.com/Supply+Chain+Attacks/Models/Using+Keras+Lambda+Layers

https://5stars217.github.io/2023-08-08-red-teaming-with-ml-models/

https://splint.gitbook.io/cyberblog/security-research/tensorflow-remote-code-execution-with-malicious-model

https://thenewstack.io/how-to-reduce-the-hallucinations-from-large-language-models/

https://aclanthology.org/2020.acl-main.249/

{   "url": "https://example.com",

"url": "https://example.com",

[Credential-related source omitted]

- [ ] 可读 ASCII/UTF-8 文本 - [ ] 看起来像路径（/xxx/yyy.php） - [ ] 看起来像 URL（http://...）

- [ ] 看起来像路径（/xxx/yyy.php） - [ ] 看起来像 URL（http://...） - [ ] 包含 flag 格式（flag{...}, NSSCTF{...}）

- [ ] 看起来像 URL（http://...）