Uploaded April 2026 | Updated September 2026, 2 weeks ago
As AI agents gain autonomy through the Model Context Protocol (MCP), they increasingly access internal APIs, secrets, and systems—often beyond traditional security visibility or control. These agent-to-agent (A2A) interactions introduce new reconnaissance blind spots and exposure pathways inside modern Zero Trust environments. Building on continued research, this session introduces the MCP Deception Incubator—a deception-as-a-framework approach designed to detect rogue or unintended AI behavior at the reconnaissance stage. Built on serverless edge workers and open deception primitives such as Canarytokens, the framework enables defenders to deploy no-cost, high-fidelity MCP honeytraps across multiple surfaces, including APIs, DNS, kubeconfigs, and credentials—without operational overhead. These traps integrate with MCP Gateways or AI integration layers that broker communication between AI agents and organizational systems, enabling early, metadata-rich visibility into autonomous activity. When an AI agent interacts with a decoy endpoint, the resulting telemetry exposes its access path, tool awareness, and decision context—transforming passive reconnaissance into actionable intelligence. Through a live demonstration, attendees will see how edge-based traps trigger in real time and how alerts integrate with SOC pipelines for correlation, credential rotation, and sustained visibility. The session then unpacks the framework’s architecture, showing how deception can be operationalized within MCP-based or agent-driven environments. Finally, the talk reframes deception from isolated honeypots into an orchestrated Zero Trust signal layer, enabling organizations to transform AI curiosity into proactive defense. Attendees will leave with reference architectures, deployment patterns, and practical guidance to embed deception into their own AI-security workflows.
Harshad Sadashiv Kadam
Indeed
Senior Security Infrastructure Engineer
Austin, Texas
harshadsadashivkadam.com
Harshad Sadashiv Kadam is a Senior Security Infrastructure Engineer at Indeed with over a decade of experience driving Zero Trust, SASE, and multi-cloud security for global enterprises. He thrives at the intersection of strategy and hands-on engineering, designing next-generation defenses against AI-driven threats while mentoring emerging security talent. Previously a Cloud Infrastructure Engineering Manager, Harshad led global cloud migrations and large-scale reliability initiatives, delivering multimillion-dollar efficiencies and strengthening operational resilience. As a mentor and AI-Champions Guild leader, he is passionate about fostering inclusion and inspiring the next generation of security innovators.
-
Managed by the OWASP® Foundation
owasp.org
As AI agents gain autonomy through the Model Context Protocol (MCP), they increasingly access internal APIs, secrets, and systems—often beyond traditional security visibility or control. These agent-to-agent (A2A) interactions introduce new reconnaissance blind spots and exposure pathways inside modern Zero Trust environments. Building on continued research, this session introduces the MCP Deception Incubator—a deception-as-a-framework approach designed to detect rogue or unintended AI behavior at the reconnaissance stage. Built on serverless edge workers and open deception primitives such as Canarytokens, the framework enables defenders to deploy no-cost, high-fidelity MCP honeytraps across multiple surfaces, including APIs, DNS, kubeconfigs, and credentials—without operational overhead. These traps integrate with MCP Gateways or AI integration layers that broker communication between AI agents and organizational systems, enabling early, metadata-rich visibility into autonomous activity. When an AI agent interacts with a decoy endpoint, the resulting telemetry exposes its access path, tool awareness, and decision context—transforming passive reconnaissance into actionable intelligence. Through a live demonstration, attendees will see how edge-based traps trigger in real time and how alerts integrate with SOC pipelines for correlation, credential rotation, and sustained visibility. The session then unpacks the framework’s architecture, showing how deception can be operationalized within MCP-based or agent-driven environments. Finally, the talk reframes deception from isolated honeypots into an orchestrated Zero Trust signal layer, enabling organizations to transform AI curiosity into proactive defense. Attendees will leave with reference architectures, deployment patterns, and practical guidance to embed deception into their own AI-security workflows.
Harshad Sadashiv Kadam
Indeed
Senior Security Infrastructure Engineer
Austin, Texas
harshadsadashivkadam.com
Harshad Sadashiv Kadam is a Senior Security Infrastructure Engineer at Indeed with over a decade of experience driving Zero Trust, SASE, and multi-cloud security for global enterprises. He thrives at the intersection of strategy and hands-on engineering, designing next-generation defenses against AI-driven threats while mentoring emerging security talent. Previously a Cloud Infrastructure Engineering Manager, Harshad led global cloud migrations and large-scale reliability initiatives, delivering multimillion-dollar efficiencies and strengthening operational resilience. As a mentor and AI-Champions Guild leader, he is passionate about fostering inclusion and inspiring the next generation of security innovators.
-
Managed by the OWASP® Foundation
owasp.org
