Mitigations
10 course-of-action (mitigation) STIX objects addressing adversary behaviors in LLM-integrated IDEs.
AI Configuration File Integrity Monitoring
Implement file integrity monitoring and diff analysis for AI configuration files (.cursorrules, .github/copilot-instructions, MCP configs). Flag non-obvious instruction content and enforce review requirements for AI configuration changes in repositories.
AI Network Traffic Segmentation
Implement egress filtering restricting LLM API and MCP tool server traffic to approved endpoints. Baseline normal LLM API traffic patterns and alert on deviations. Restrict agent-initiated outbound connections to pre-approved destinations.
Agent Command Allowlisting
Implement command allowlists restricting agent-executable commands to development-relevant operations. Block system enumeration commands, cloud metadata access, and administrative operations unless explicitly approved per-task.
Agent Execution Sandboxing
Run AI coding agents in isolated security contexts with least-privilege permissions separate from the developer's ambient session. Implement task-scoped permission grants that restrict agent capabilities to files and tools relevant to the current task.
Context Window Content Filtering
Apply input sanitization and prompt injection detection to content entering the LLM context window. Scan for instruction-like patterns in code comments, documentation, and external content. Implement content trust levels differentiating project files from external sources.
Credential Isolation from AI Agents
Prevent AI agent processes from accessing the developer's credential stores, SSH key directories, cloud configuration files, and authentication tokens. Use credential proxies that provide task-scoped, time-limited access.
Extension Security Controls
Enforce extension allowlisting from verified publishers. Flag extensions requesting LLM API access combined with network permissions. Monitor extension API calls for prompt/response interception. Restrict sideloading from non-marketplace sources.
Generated Code Security Scanning
Apply inline SAST/security scanning to AI-generated code before presentation to the developer. Track vulnerability detection rates over time to identify adversarial steering patterns. Block acceptance of code with known vulnerability patterns.
LLM Output Validation and Encoding Detection
Scan LLM-generated output for encoded data patterns (base64, URL encoding), embedded URLs, and content that diverges from the prompt intent. Implement output content policies that block exfiltration patterns in generated code, markdown rendering, and tool invocations.
MCP Server Allowlisting and Verification
Maintain an approved inventory of MCP tool servers. Require signature verification for server registration. Validate that registered endpoints match approved providers. Alert on new server registrations from project-level configuration.