AISecHub

1,962 posts
Opens profile photo
AISecHub
@AISecHub
🚀 AISecHub | AI & Cybersecurity | Securing AI systems, and sharing insights on emerging challenges 💡Schedule a meeting calendly.com/innovguard/mee
Singaporelinkedin.com/groups/1454551…
Joined December 2024
7,264 Following
7,884 Followers

AISecHub’s posts

AISecHub
@AISecHub
26 Awesome AI Security Lists 1️⃣ github.com/wearetyomsmnv/ 2️⃣ github.com/kaplanlior/oss 3️⃣ github.com/corca-ai/aweso 4️⃣ github.com/zmre/awesome-s 5️⃣ github.com/DeepSpaceHarbo 6️⃣ github.com/Billy1900/Awes 7️⃣ github.com/trailofbits/aw 8️⃣ github.com/RiccardoBiosas 9️⃣ github.com/disesdi/mlseco 🔟
Image
27K
AISecHub
@AISecHub
Benchmarking AI Agents: 175 Tasks in a Securely Designed Testing Environment The study "Benchmarking LLM Agents on Consequential Real-World Tasks" evaluates AI systems' ability to autonomously handle professional workflows. Conducted across 175 tasks, the benchmark tests AI
Image
5.3K
AISecHub
@AISecHub
Great share by
@olivier_boschko
from
@dreadnode
!
Quote
Tal Eliyahu
@Eliyahu_Tal_
Adversarial Machine Learning Challenges: A Deeper Dive @olivier_boschko from @dreadnode explores the evolving field of adversarial machine learning (#AML), focusing on techniques that manipulate inputs to mislead models without perceptible changes to their appearance
Image
438
AISecHub
@AISecHub
AI-Driven Security Research: Weekly Highlights 🔍 This week’s 16 studies explore advancements in hate speech detection, adaptive security for LLMs, code authorship attribution, adversarial manipulations, secure software engineering, and smart contract analysis. 😡 AfriHate: A
Image
12K
AISecHub
@AISecHub
🔍 AI Security & Compliance: Risk Management in High-Risk AI Systems The High-Risk EU AI Act Toolkit (HEAT) provides a structured methodology for ensuring compliance with the EU Artificial Intelligence Act. One of the key focus areas is the risk management system, which mandates
Image
4.7K
AISecHub
@AISecHub
🛡️ The 5 Security Levels for AI Systems
Quote
Tal Eliyahu
@Eliyahu_Tal_
🛡️ The 5 Security Levels for AI Systems The @RANDCorporation report introduces a 5-level security framework (SL1 to SL5) to guide organizations in assessing and implementing security measures against escalating threats to AI systems, with a strong emphasis on protecting AI model
Image
Image
Image
4.5K
AISecHub
@AISecHub
The paper "Emerging Security Challenges of Large Language Models" examines the vulnerabilities of Large Language Models (LLMs), focusing on adversarial attacks, data poisoning, and risks associated with their complex supply chains. It highlights how the reliance on vast,
Image
6.7K
AISecHub
@AISecHub
🔍 AI Security Weekly — Feb 16, 2025 ⚫ We Need to Integrate and Unify for AI Security An interesting perspective from Sven Cattell (
@comathematician
) on adopting a CWE-like system for AI security, featuring transparent disclosures — similar to CVE processes — to balance
Image
1.8K
AISecHub
@AISecHub
31
@awscloud
Key Threat Scenarios for Generative AI Workloads (by
@Eliyahu_Tal_
) docs.google.com/spreadsheets/d #AISecurity #Cybersecurity #AITrust #AIRegulation #AIRisk #AISafety #LLMSecurity #ResponsibleAI #DataProtection #AIGovernance #AIGP #SecureAI #AIAttacks #AICompliance
docs.google.com
31 AWS Key Threat Scenarios for Generative AI Workloads (@AISecHub).xlsx
5.2K
AISecHub
@AISecHub
Awesome List of Cybersecurity and AI by Kris Oosthoek (
@f00th0ld
) A curated resource exploring key topics in AI and cybersecurity, including AI embeddings security, model attacks, LLM-based threats, AI policy frameworks, defensive/offensive AI use cases, academic research, and
5.8K
AISecHub
@AISecHub
OWASP Top 10 for LLM Applications Mindmap 📊 This mindmap explores the relationships within the OWASP Top 10 for LLM Applications, while drawing comparisons with the OWASP Top 10 for Web Applications and MITRE ATLAS. Author: Henry Hu, Auriga Security, Inc.,
@owasp
#Taiwan
Image
1.3K
AISecHub
@AISecHub
📄 Toward Intelligent and Secure Cloud: Large Language Model Empowered Proactive Defense The Challenge ⚠️ The rapid evolution of cloud computing technologies and the increasing number of cloud applications have provided significant benefits in daily lives. However, the
Image
2K
AISecHub
@AISecHub
Interesting,
@andyzou_jiaming
!
Quote
Andy Zou
@andyzou_jiaming
Robot dogs were jailbroken to deliver explosives. Self-driving LLM was hacked to target pedestrians. This may be one of the last warnings before real-world AI agents cause significant harm. Many downplay LLM jailbreaks, comparing them to Google searches. But when LLMs control x.com/AlexRobey23/st…
482
AISecHub
@AISecHub
AI Security Solution Cheat Sheet Q1-2025 by
@owasp
The LLM and Gen Ai Security Solutions Guide Cheat Sheets provide a quick way to view and reference guidance and resources from the OWASP LLM and Gen AI Security Landscape guide. This edition includes Cheat Sheets for the
Image
7K
AISecHub
@AISecHub
Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attacks Integrating LLM agents with memory modules, APIs, and web access enhances functionality but introduces new vulnerabilities. Adversaries can exploit these through memory poisoning, adversarial web
Image
10K
AISecHub
@AISecHub
Interesting post by Marcin Niemiec (
@xvnpw
)
Quote
Marcin Niemiec
@xvnpw
🔒 New blog post: "Scaling Threat Modeling with AI: Generating 1000 Threat Models Using Gemini 2.0" I used Google's Gemini to automate security documentation at scale. Check out: 📝 Blog: xvnpw.github.io/posts/scaling- 🛠️ Code: github.com/xvnpw/sec-docs
1.2K
AISecHub
@AISecHub
🧠 Exploring AI Security Through the Shared Responsibility Model The shared responsibility model, introduced by
@Microsoft
, emphasizes the importance of collaboration between providers and users to address the evolving challenges in AI security. By defining clear roles,
Image
3.6K