Lightning Talks: Detecting Data Exfiltration on the Edge with Pixie
summary
Pixie uses eBPF to collect request bodies and system events from every node without changing apps, running analysis locally via a Python/Pandas-like scriptable interface to flag sensitive fields. A demo traces mixed HTTP/HTTPS egress and surfaces a request leaking name, card number and phone, and the agent targets under 5% overhead (typical 2–3%). Deployable per node as in-cluster edge compute, it supports zero instrumentation, captures encrypted traffic, and keeps query and results inside the cluster or on-prem.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:45Disclaimer
- 1:08What is data exfiltration risk
- 1:59What is Pixie
- 3:30Pixie Architecture
- 4:05Pixie Language
- 4:25How to Catch Data Leaks
- 5:05Demo Scenario
- 5:34Pixie Demo
related talks
26mTriaging Real-time Security Threats with eBPF-powered Observability
14mFlexible, Open and Easy Observability for Developers
30mData Science for Infrastructure: Observe, Understand, Automate
34mTotal Visibility: Harnessing Pixie for a Full Observability Landscape
04mInstant, Auto-telemetry for K8s Observability
17meBPF in Microservices Observability
28mBpftrace Meets Pixie: Dynamic Monitoring of K8s Clusters Unleashed
29mReproducing Production Issues in your CI Pipeline Using eBPF