Security Considerations for Large Language Model Use: Implementation Research in Securing LLM-Integrated Applications
Article Sidebar
Main Article Content
Abstract
Large Language Models (LLMs) are rapidly being adopted in various applications due to their natural language capabilities that enable user interaction using human language. As system designers, developers, and users embrace generative artificial intelligence and large language models in various applications, they need to understand the significant security risks associated with them. The paper describes a typical LLM-integrated application architecture and identifies multiple security risks to address while building these applications. In addition, the paper provides guidance on potential mitigations to consider in this rapidly evolving space to help protect systems and users from potential attack vectors. This paper presents the common real-world application patterns of LLMs trending today. It also provides a background on generative artificial intelligence and related fields.
Downloads
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
How to Cite
References
Chatgpt, openai.com/chatgpt/.
CHATGPT Sets Record for Fastest-Growing User Base - Analyst Note | Reuters, www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/.
“Command R.” Cohere, cohere.com/models/command.
Meta Llama, ai.meta.com/llama/.
GPT-4, openai.com/research/gpt-4/.
McGraw, Gary, et al. An Architectural Risk Analysis of Large Language Models: Applied Machine Learning Security, berryvilleiml.com/docs/BIML-LLM24.pdf.
Artificial Intelligence Definitions, hai.stanford.edu/sites/default/files/2020-09/AI-Definitions-HAI.pdf.
“Generative AI Foundations: An Introduction to Basic Generative AI Concepts.” Sendbird, sendbird.com/developer/tutorials/introduction-to-basic-generative-ai-concepts.
“Introduction to Deep Learning.” GeeksforGeeks, 26 May 2024, www.geeksforgeeks.org/introduction-deep-learning/.
Team, Toloka. Difference between AI, ML, LLM, and Generative AI, 23 May 2024, toloka.ai/blog/difference-between-ai-ml-llm-and-generative-ai/.
“What Are Large Language Models?: Nvidia Glossary.” NVIDIA, www.nvidia.com/en-us/glossary/large-language-models/.
Sajid, Haziqa. “A Comprehensive Overview of Large Language Models.” Wisecube, 1 June 2023, www.wisecube.ai/blog/a-comprehensive-overview-of-large-language-models/.
OWASP, owasp.org/www-project-top-10-for-large-language-model-applications/assets/PDF/OWASP-Top-10-for-LLMs-2023-v1_1.pdf.
Smith, Craig S. “What Large Models Cost You – There Is No Free AI Lunch.” Forbes, Forbes Magazine, 20 Feb. 2024, www.forbes.com/sites/craigsmith/2023/09/08/what-large-models-cost-you--there-is-no-free-ai-lunch/?sh=44cc5add4af7.
McGuinness, Patrick. “GPT-4 Details Revealed.” GPT-4 Details Revealed - by Patrick McGuinness, AI Changes Everything, 12 July 2023, patmcguinness.substack.com/p/gpt-4-details-revealed.
“LLM Training: A Simple 3-Step Guide You Won’t Find Anywhere Else!” Medium, Medium, 10 Mar. 2024, masteringllm.medium.com/llm-training-a-simple-3-step-guide-you-wont-find-anywhere-else-98ee218809e5.
“Fine-Tuning Large Language Models (Llms) in 2024.” SuperAnnotate, www.superannotate.com/blog/llm-fine-tuning.
Ben Dickson, et al. “The Complete Guide to LLM Fine-Tuning.” TechTalks, 15 Aug. 2023, bdtechtalks.com/2023/07/10/llm-fine-tuning/.
“Security Guardrails for LLM: Ensuring Ethical AI Deployments.” Security Guardrails for LLM: Ensuring Ethical AI Deployments, www.turing.com/resources/implementing-security-guardrails-for-llms.
“LLMS Guardrails Guide: What, Why & How: Attri AI Blog: Attri.Ai Blog.” LLMs Guardrails Guide: What, Why & How | Attri AI Blog | Attri.Ai Blog, attri.ai/blog/a-comprehensive-guide-everything-you-need-to-know-about-llms-guardrails.
Daryanani, Lavina. “How to Jailbreak Chatgpt.” Watcher Guru, 7 Feb. 2023, watcher.guru/news/how-to-jailbreak-chatgpt.
Greshake, Kai, et al. “Not What You’ve Signed up for Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection.” arXiv.Org, 5 May 2023, doi.org/10.48550/arXiv.2302.12173. https://doi.org/10.1145/3605764.3623985
Liu, Yi, et al. “Prompt Injection Attack against LLM-Integrated Applications.” arXiv.Org, 2 Mar. 2024, doi.org/10.48550/arXiv.2306.05499.
Zou, Andy, et al. “Universal and Transferable Adversarial Attacks on Aligned Language Models.” arXiv.Org, 20 Dec. 2023, doi.org/10.48550/arXiv.2307.15043.
“CHATGPT Plugin Exploit Explained: From Prompt Injection to Accessing Private Data · Embrace the Red.” Embrace The Red, 28 May 2023, embracethered.com/blog/posts/2023/chatgpt-cross-plugin-request-forgery-and-prompt-injection./.
Xu, Ziwei, et al. “Hallucination Is Inevitable: An Innate Limitation of Large Language Models.” arXiv.Org, 22 Jan. 2024, doi.org/10.48550/arXiv.2401.11817.
Garcia, Marisa. “What Air Canada Lost in ‘Remarkable’ Lying AI Chatbot Case.” Forbes, Forbes Magazine, 19 Feb. 2024, www.forbes.com/sites/marisagarcia/2024/02/19/what-air-canada-lost-in-remarkable-lying-ai-chatbot-case/.
Russell, Josh. “Judge Won’t Sanction Michael Cohen over Ai-Generated Fake Legal Cases.” Courthouse News Service, 20 Mar. 2024, www.courthousenews.com/judge-wont-sanction-michael-cohen-over-ai-generated-fake-legal-cases/.
Mauran, Cecily. “Whoops, Samsung Workers Accidentally Leaked Trade Secrets via Chatgpt.” Mashable, Mashable, 6 Apr. 2023, mashable.com/article/samsung-chatgpt-leak-details.
Huynh, Daniel. “POISONGPT: How to Poison LLM Supply Chainon Hugging Face.” Mithril Security Blog, Mithril Security Blog, 18 Dec. 2023, blog.mithrilsecurity.io/poisongpt-how-we-hid-a-lobotomized-llm-on-hugging-face-to-spread-fake-news/.
Knapton, Ken. “Council Post: Navigating the Biases in LLM Generative AI: A Guide to Responsible Implementation.” Forbes, Forbes Magazine, 5 Oct. 2023, www.forbes.com/sites/forbestechcouncil/2023/09/06/navigating-the-biases-in-llm-generative-ai-a-guide-to-responsible-implementation/.
Weng, Lilian. “LLM Powered Autonomous Agents.” Lil’Log (Alt + H), 23 June 2023, lilianweng.github.io/posts/2023-06-23-agent/.
Thiel, David. “Investigation Finds AI Image Generation Models Trained on Child Abuse.” FSI, cyber.fsi.stanford.edu/news/investigation-finds-ai-image-generation-models-trained-child-abuse.
Crist, Carolyn. “Chatgpt Gives Incorrect Answers about Eoe.” Medscape, 16 Nov. 2023, www.medscape.com/viewarticle/998537?form=fpf.
Hague, Danny. “Multimodality, Tool Use, and Autonomous Agents: Large Language Models Explained, Part 3.” Center for Security and Emerging Technology, 25 Mar. 2024, cset.georgetown.edu/article/multimodality-tool-use-and-autonomous-agents/.
Ansari, M. Z., Ahmad, T., & Fatima, A. (2019). Feature Selection on Noisy Twitter Short Text Messages for Language Identification. In International Journal of Recent Technology and Engineering (IJRTE) (Vol. 8, Issue 4, pp. 10505–10510). https://doi.org/10.35940/ijrte.d4360.118419
Lalaei, R. A., & Mahmoudabadi, Dr. A. (2024). Promoting Project Outcomes: A Development Approach to Generative AI and LLM-Based Software Applications’ Deployment. In International Journal of Soft Computing and Engineering (Vol. 14, Issue 3, pp. 6–13). https://doi.org/10.35940/ijsce.d3636.14030724
Rao P, Mr. V., & Sivakumar, Dr. A. P. (2020). A Comprehensive Retrospection of Literature Reported Works of Community Question Answering Systems. In International Journal of Innovative Technology and Exploring Engineering (Vol. 9, Issue 3, pp. 1904–1907). https://doi.org/10.35940/ijitee.c8769.019320