APEX’s Post

#ai security is as fast a changing area as any in #cybersecurity. It's also one that's yet to be underpinned by a large body of research, which means that there are a range of sometimes conflicting viewpoints out there on the topic. For example, on the one hand some folk suggest that "There's not a lot of magical incantations that you can send to an LLM and have it spit out passwords and sensitive info... But there's a lot of vulnerabilities in the servers that are used to host LLMs. The [LLM] is really not where you're going to get hacked — you're going to get hacked from all the tools you use around the LLM"... and on the other people talk about the risks of algorithm and dataset poisoning as a way to hack the LLM itself. Which perspective is correct...? They both are, which is why it's important to treat #AI like every other major system and perform a complete security risk assessment and security architecture design before implementing an AI solution, and then to actively test, monitor and maintain your security controls throughout its lifecycle. https://lnkd.in/g_5K6Kfe

As we continue to marvel at AI's potential to transform our lives , it's all too easy to overlook the importance of safeguarding our data and privacy. This comprehensive piece by Frédéric A. serves as a crucial reminder that security should be a fundamental part of creating and using AI technologies, not an afterthought. #AI #dataprivacy #cybersecurity #azure #azureopenai

Security Best Practices for GenAI Applications Large Language Models (LLMs) hold great potential but also pose significant security challenges that need to be addressed by developers and administrators of GenAI applications. These challenges involve preserving data confidentiality and integrity throughout the training and querying of LLMs, guaranteeing their availability and reliability, addressing potential misuse by malicious entities, monitor and more. This article covers a few domains: ☆ Data Security ☆ Network Security ☆ Access and Identity Security ☆ Application Security ☆ Governance Security #genai #Azure #openai #security #bestpractices

Great overview of GenAI sec in this write up from VentureBeat. TLDR: “gen AI doesn’t fit traditional security paradigms focused on preventing data from going to people who aren’t supposed to have it.” #AI #cybersecurity #databreach.

The article discusses vulnerabilities in AI-as-a-Service (AIaaS) platforms, specifically highlighting Hugging Face. For more information, you can read the full article here: https://lnkd.in/d_BEzgsy #AI #cybersecurity #vulnerability

#ai #security Are you building LLM-based products or apps? This repo has the most comprehensive description of how LLMs can be attacked (particularly with prompt injection) that I've seen so far. It describes the potential consequences of giving LLMs access to influence other applications and describes these 7 attack scenarios: - Remote control of LLMs - Leaking/exfiltrating user data - Persistent compromise across sessions - Spread injections to other LLMs - Compromising LLMs with tiny multi-stage payloads - Automated Social Engineering - Targeting code completion engines Check it out 👇 https://lnkd.in/gsp-zMWy #llm #saas #cybersecurity #ResilientSoftwareSecurity

First Step in AI/ML Security Is Finding Them All - https://buff.ly/48B8jtc #AI #ML #security #AIsecurity #cybersecurity #infosec #ITsecurity

You cannot secure something you do know you have. How many companies have 100% accurate inventory of AI/ML being used that stays current? What about vendors that use AI/ML and may be training or at least accessing corporate data? https://lnkd.in/eCZcF_5Q #ai #security #cybersecurity

Knowing how to implement AI at scale in our organizations is a business imperative, doing it in a secure manner is a must. Learn Best Practices of Security for GenAI shared by Antonio Formato: ✔ 𝐃𝐚𝐭𝐚 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 ✔𝐍𝐞𝐭𝐰𝐨𝐫𝐤 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 ✔𝐀𝐜𝐜𝐞𝐬𝐬 𝐚𝐧𝐝 𝐈𝐝𝐞𝐧𝐭𝐢𝐭𝐲 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 ✔𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 ✔𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 https://lnkd.in/dWxRG-pn

Security Best Practices for GenAI Applications Large Language Models (LLMs) hold great potential but also pose significant security challenges that need to be addressed by developers and administrators of GenAI applications. These challenges involve preserving data confidentiality and integrity throughout the training and querying of LLMs, guaranteeing their availability and reliability, addressing potential misuse by malicious entities, monitor and more. This article covers a few domains: ☆ Data Security ☆ Network Security ☆ Access and Identity Security ☆ Application Security ☆ Governance Security #jsisen #phuongit #viettechgroupvn #genai #Azure #openai #security #bestpractices https://lnkd.in/gZFa3NZx