2.2 KiB
2.2 KiB
Runtime Instrumentation
Bridges eBPF-based runtime monitoring into the Stella Ops platform, converting kernel-level events into canonical format for reachability validation and signal scoring.
Purpose
Runtime Instrumentation adapts raw eBPF events from Tetragon into the Stella Ops canonical RuntimeCallEvent format. This enables the platform to incorporate live runtime observations (system calls, function probes, process lifecycle) into reachability validation and evidence-weighted vulnerability scoring without coupling downstream modules to any specific eBPF agent.
Quick Links
- Architecture - Technical design and implementation details
Status
| Attribute | Value |
|---|---|
| Maturity | Beta |
| Source | src/RuntimeInstrumentation/ |
Key Features
- Tetragon gRPC client: Connects to the Tetragon agent's gRPC stream and ingests raw eBPF events in real time
- eBPF probe type mapping: Supports all major probe types -- Kprobe, Kretprobe, Uprobe, Uretprobe, Tracepoint, USDT, Fentry, Fexit, ProcessExec, ProcessExit
- Stack frame canonicalization: Converts raw kernel/user-space stack frames into
CanonicalStackFramewith symbol resolution and address normalization - Hot symbol index updates: Publishes observed symbols to the hot symbol index for runtime reachability correlation
- Privacy filtering: Strips sensitive data (environment variables, command arguments, file paths) before events leave the instrumentation boundary
Dependencies
Upstream (this module depends on)
- Tetragon - External eBPF agent providing kernel-level event streams via gRPC
Downstream (modules that depend on this)
- Signals - Consumes
RuntimeCallEventdata for runtime signal scoring (RTS dimension) - Scanner - Uses runtime observations for reachability validation
- Policy - Incorporates runtime evidence into policy evaluation and verdicts
Related Documentation
- Signals - Runtime signal scoring using RTS dimension
- Signals eBPF Contract - Determinism profile for eBPF witnesses
- Scanner - Reachability validation
- Policy - Runtime evidence in policy decisions