FraudGPT writes fileless malware, OWASP Top 10 LLM v1.0, LLM Backdoor attacks, Nvidia publish prompt injection code and mitigation techniques
ProCheckUp (PCU)
Highly accredited independent cyber security expertise. Find out why clients give PCU a 96/100 rating.
Hey Security Experts! Welcome to another edition of PCU's Generative AI Security Newsletter, the free newsletter that helps you keep up with the advances and impacts of cyber security relating to Generative AI tools and large language models.
The Generative AI Security / LLM Security industry continues to progress at a lightning-fast pace. This week, OWASP released version 1.0 of their Top 10 for LLM, providing guidance on the key threats for LLM Applications. NVidia's LLM Red Team provides detailed guidance on how to exploit Langchain-based applications, as well as how to mitigate prompt-injection attacks. FraudGPT comes to the forefront as a competitor to WormGPT, enabling LLM-powered offensive hacking and OSINT with a $1,600 / year license.
Let's explore the most relevant news of the last week:
Generative AI Security News
?? OWASP Top 10 for LLM - Version 1.0 is released
https://www.dhirubhai.net/pulse/official-release-owasp-top-10-large-language-model-v10-steve-wilson/
?? Nvidia - How to start securing LLM apps against prompt injection attacks
?? Boring AppSec - Framework for safe LLM usage in businesses
?? Backdoor attacks vs LLMs, how to protect BERT models
?? Prompt injection for your CV - make AI LLM believe you are perfect candidate
?? FraudGPT - writes?fileless malware, build phishing pages, spot vulnerabilities
?? If you want to talk to PCU about Generative AI Testing: Contact PCU here
?
Don't forget to leave a comment below to extend this conversation.
??♀?????????♂??????? Servant Leader ? Queen's Award winning Innovator Quantum + AI ? Chairman: QBRAIN++ ResiLux 4IR:PEX Quantum Innovation Labs QuantumDiamond TriGold ATCA5000 Philanthropia mi2g?
1 年HOW DOES NVIDIA COMPETE WITH QUANTUM COMPUTERS AS AI MIGRATES TO QUANTUM? 1. Google's new Quantum Computer completes a task in 6 seconds that would have taken one of the world's fastest super computers 47 years. 2. This represents a speed that's 240+ million times greater than its 2019 model and 247 million times faster than a classical super computer. 3. Previously, Google's quantum computer performed a task in 200 seconds that would have taken the world's fastest supercomputer >10,000 years! https://www.dhirubhai.net/posts/dkmatai_qc-ai-quantumcomputing-activity-7101510218647261184-IymV?utm_source=share&utm_medium=member_android #AI #QuantumComputing #Quantum
Cyber Security & Artificial Intelligence
1 年Thanks to the following people for their research and insights into Generative AI Security: Steve Wilson, OWASP? Foundation & the OWASP LLM contributors - OWASP Top Top 10 LLM v1.0 release Rich Harang & NVIDIA - Securing LLMs against prompt injection Sandesh Mysore Anand - BoringAppSec's sage LLM usage framework Xuanli He, Jun Wang, Benjamin Rubinstein, Trevor Cohn - Protecting BERT models from insertion-based backdoor attacks Kai Greshake - Prompt Injection for your CV And all other contributors ??