P 40 | Agentic Security - Grant McCracken @ DarkHorse.sh

SummaryIn this episode, Philippe Trounev and Grant McCracken discuss the evolving landscape of cybersecurity, particularly focusing on the role of agentics and AI in enhancing security measures. They explore the implications of new attack vectors like prompt injection, the potential for agent-to-agent attacks, and the innovative approach of agentic pen testing. Grant shares insights on how AI can improve the efficiency and effectiveness of security testing while emphasizing the importance of foundational cybersecurity practices. The conversation highlights the balance between leveraging advanced technologies and maintaining robust security protocols.TakeawaysCybersecurity is evolving with the introduction of agentics.Fundamentals of cybersecurity remain crucial despite new technologies.68% of data breaches originate from human error.Prompt injection is a significant new attack vector.AI can identify existing vulnerabilities but not create new ones.Agentic pen testing offers continuous, fatigue-free testing.Contextual vulnerabilities are better identified by AI.AI pen testing could reduce costs compared to human testers.Dark Horse aims to make cybersecurity solutions accessible.AI compute costs are high but expected to decrease over time.TitlesThe Future of Cybersecurity: Agentics and AIUnderstanding Cybersecurity in the Age of AISound bites"68% of data breaches start at the human layer.""Prompt injection is a new attack vector.""Agentic pen testing can run 24/7 without fatigue."Chapters00:00 Introduction to Cybersecurity and AI Agents01:42 Changing Attack Vectors in Cybersecurity03:44 Human Layer Vulnerabilities and AI's Role07:13 Prompt Injection and Offensive Strategies11:26 Agentic Systems and Security Risks12:58 Agent-to-Agent Attacks and Communication Risks17:03 Agentic Pen Testing: A New Approach21:33 Cost Efficiency of Agentic Pen Testing24:48 Exciting Developments in AI Pen Testing27:53 Future of Agentic Pen Testing and Market Potential

There are no comments yet.

Please log in to write the first comment.