๐ช๐ฒ๐ฏ๐ถ๐ป๐ฎ๐ฟ ๐๐น๐ฒ๐ฟ๐ | ๐๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐ฆ๐ฒ๐ฟ๐ถ๐ฒ๐ #๐ฎ: ๐ง๐ต๐ฟ๐ฒ๐ฎ๐ ๐ ๐ผ๐ฑ๐ฒ๐น๐น๐ถ๐ป๐ด ๐ถ๐ป ๐๐ ๐ฆ๐๐๐๐ฒ๐บ๐
Securing the Invisible Thread: Threat Modeling and Defending MCP in Agentic AI Systems. As agentic AI adoption accelerates, most security discussions still focus on the model. MCP connects LLMs to tools, APIs, data, and systems. In doing so, it quietly reshapes trust boundaries and expands the attack surface beyond what traditional AppSec models were built for.
๐ ๐๐ฎ๐๐ฒ: 31st January 2026 (Saturday) โฐ ๐ง๐ถ๐บ๐ฒ: 8:00 PM IST Link - https://www.linkedin.com/uas/login?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Fevents%2Failearningseries-2-threatmodell7422121047816257538%2F
In this talk, I explore how MCP can turn well-intentioned agents into effective attack orchestrators through: โข context poisoning โข token and secret leakage โข privilege escalation via scope creep โข tool misuse โข shadow MCP servers
The goal isnโt to slow AI innovation, itโs to secure how power flows.
#AISecurity #AgenticAI #MCP #ThreatModeling #AppSec #OWASP #LLMSecurity #GenAI #AITrust MoS