Skip to content

Phase 1 Summary

Status: ✅ Completed Focus: Linux Kernel Hardening, Anomaly Detection Theory, and Prototype v0.7.

This phase focused on establishing the “Ground Truth” for the Sentinel environment. Before we can detect anomalies, we must mathematically define what a “normal” Linux process looks like.


We successfully engineered a raw C interceptor using ptrace that demonstrates Active Defense:

  • Mechanism: PTRACE_SYSCALL loop.
  • Detection: Identified mkdir syscalls (ID 83).
  • Response: Neutralized the call by rewriting the RAX register to -1 (Operation not permitted).

We moved away from standard Random Forests and implemented a Weightless Neural Network (WiSARD) approach:

  • Why: Traditional NN is too slow for runtime interception.
  • Method: Boolean RAM-based learning (One-Shot Learning).
  • Metric: Hamming Distance for anomaly scoring.

To ensure the researcher (us) is safe while handling malware, the lab was hardened:

  • Network: UFW Firewall automation.
  • Access: SSH Key-only auth + Fail2Ban IPS.
  • Integrity: SHA-256 File Integrity Monitoring (FIM).

For the daily experimental logs and source code, refer to the Systems Security Foundations Repository.