01 / CONTEXT
The problem
Indicators of compromise on a Windows machine live scattered across Event Viewer, active processes and network connections. Catching an anomaly in time means crossing all of that by hand.
Security observability · 2026
A local observability hub that turns scattered system signals into clear, auditable decisions.
A local Python daemon that audits processes via WinAPI, correlates suspicious signals and translates them into a single panel with severity and context—nothing leaves the machine.
01 / CONTEXT
Indicators of compromise on a Windows machine live scattered across Event Viewer, active processes and network connections. Catching an anomaly in time means crossing all of that by hand.
02 / DIRECTION
A local Python daemon that audits processes via WinAPI, correlates suspicious signals and translates them into a single panel with severity and context—nothing leaves the machine.
03 / EXECUTION
04 / RESULT
05 / IMPLEMENTATION
import ctypes
import win32process
import win32con
def audit_running_processes():
"""Scan process memory surfaces for anomalous handles."""
print("[*] Initializing CyberSentinel Local Observability...")
anomalies = []
# Query system information via WinAPI
processes = win32process.EnumProcesses()
for pid in processes:
try:
h_proc = ctypes.windll.kernel32.OpenProcess(
win32con.PROCESS_QUERY_INFORMATION | win32con.PROCESS_VM_READ,
False, pid
)
if h_proc:
# Audit handle safety signature
sig = check_security_signature(h_proc)
if not sig:
anomalies.append(pid)
ctypes.windll.kernel32.CloseHandle(h_proc)
except Exception:
continue
print(f"[+] Scan finished. Detected {len(anomalies)} process anomalies.")
return anomalies06 / LEARNING
I would separate the collection layer from the heuristics earlier: mixing them at the start made iterating on false positives slower. That is the kind of debt you only see once the system is running.