Collect
Bring relevant security events, logs, alerts, or structured data into a workflow that can be reviewed consistently.
← Back to Projects
Security Operations Research
SentinelForge
AI-Assisted Security Operations
A research project exploring how locally controlled AI can assist security operations through event classification, log summarization, incident-response workflows, and automation.
Security OperationsLocal AILog AnalysisIncident ResponseAutomationPythonPowerShellResearch
Overview
Why I'm Exploring It
Security teams often work with large amounts of repetitive, technical information. Logs, alerts, event timelines, documentation, and investigation notes can consume significant analyst time.
SentinelForge is my research environment for exploring where AI might assist those workflows without pretending that a language model can replace evidence, security tooling, or analyst judgment.
The project is deliberately positioned as research. The objective is to test useful patterns, understand limitations, and build small repeatable workflows before treating anything as an operational security capability.
Workflow
Where AI Might Assist
Classify
Explore whether local AI can help group, label, prioritize, or summarize events without replacing analyst judgment.
Investigate
Use structured prompts and supporting context to help surface relationships, questions, and possible next investigative steps.
Respond
Explore repeatable workflows that assist documentation, containment planning, communication, and follow-up actions.
Research Areas
Questions Before Automation
The first goal is understanding where AI adds useful assistance and where it adds noise, uncertainty, or unnecessary risk.
Project Status
SentinelForge is currently a research project. Planned capabilities on this page represent areas for experimentation, not production security controls.
Event Classification
Explore whether local models can help categorize security events into useful buckets for analyst review.
Log Summarization
Condense large amounts of technical event data into shorter summaries while preserving the original evidence for validation.
Incident Assistance
Use AI as a supporting tool for investigative questions, documentation, timelines, and response checklists.
Python Automation
Develop small Python utilities for processing structured data, normalizing inputs, and connecting parts of the workflow.
PowerShell Automation
Explore repeatable Windows and Microsoft security administration tasks that can be safely scripted.
Local Model Testing
Compare model behavior across security prompts while keeping sensitive lab data inside locally controlled infrastructure.
Guardrails
AI Assistance Still Needs Evidence
Human Validation
Model output is treated as assistance that requires verification rather than authoritative security evidence.
Source Preservation
Original logs and event data remain available so summaries can always be checked against the underlying evidence.
Local Processing
Local models provide an opportunity to experiment with sensitive lab data without automatically sending it to external providers.
Controlled Automation
Automated actions should be narrow, understandable, reversible, and separated from unrestricted model decision-making.
“The useful question is not whether AI can make a security decision. It is where AI can reduce repetitive work while leaving evidence and judgment intact.”
Research Goals
What I Want to Learn
01
Which security tasks benefit from summarization without losing important technical context.
02
How reliably local models can classify structured security events.
03
Where automation can improve consistency without creating unsafe autonomous behavior.
04
How local AI infrastructure can support privacy-sensitive security workflows.
Next Steps
From Research to Prototype
01
Build a repeatable event-input format
02
Test classification prompts across local models
03
Prototype log-summary workflows
04
Create small Python processing utilities
05
Explore safe PowerShell automation
06
Document model limitations and failure cases
More Projects