Skip to content

mohazab2006/SentinelStack

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SentinelStack AI

SentinelStack is a Dockerized cyber security platform that combines deterministic detections with behavioral AI signals and analyst-facing AI triage.

It is built to feel like a real SOC workflow: ingest traffic, extract behavior, detect threats, fuse scores, trigger response, and provide explainable alert context.


What SentinelStack Does

  • Ingests HTTP request telemetry from your app flow
  • Detects known attack patterns with explicit rules
  • Detects suspicious unknown behavior with anomaly scoring
  • Fuses rule score + anomaly score into a clear severity outcome
  • Automates response (alerting, flagging, optional auto-block)
  • Uses OpenAI for post-detection triage summaries and recommendations
  • Monitors internal exposure changes with Port Guard

AI in SentinelStack

SentinelStack is AI-powered in two layers:

  1. Behavioral anomaly detection (in detection path)
  • Feature extraction over rolling windows per source
  • Learns what normal behavior looks like, then flags unusual patterns
  • Combines fast ML checks with optional LLM advisory signal
  • Outputs anomaly_score_norm for explainable, deterministic fusion
  1. AI triage (post-detection analyst support)
  • Alert explanation
  • Advisory risk score
  • Recommended next actions

Design policy:

  • Detection and enforcement remain deterministic and auditable
  • AI influences scoring and triage context
  • Blocking is never delegated blindly to LLM output alone

Detection Pipeline

Traffic -> Ingestion -> Behavior Features -> Rule Detection + AI Anomaly Detection -> Score Fusion -> Severity -> Response -> AI Triage

Key behavior signals include:

  • requests per minute
  • failed authentication ratio
  • unique endpoints
  • 4xx / 5xx ratios
  • request timing patterns
  • path entropy
  • suspicious payload frequency
  • auth endpoint concentration

Severity and Response Model

  • LOW: log only
  • MEDIUM: log + alert
  • HIGH: alert + flagged
  • CRITICAL: alert + auto-block

Architecture

Component Role
Next.js Dashboard SOC-style UI for events, alerts, blocks, and AI status
nginx Reverse proxy + single entrypoint
FastAPI logging-service Ingestion, rule engine, anomaly scoring, fusion, response, AI enrichment
FastAPI demo-app Traffic generator / simulation app
FastAPI portguard-service Exposure monitoring and port-delta findings
Postgres Persistent event, alert, and block data

About

AI-powered cybersecurity platform that combines deterministic threat detection, behavioral anomaly analysis, and OpenAI-assisted triage to deliver explainable alerts, severity scoring, and automated response in a full Dockerized SOC stack.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors