sajjit.soti
← Back to projects

Case study — Personal cloud security project

Self-Healing Cloud SOC: from passive monitoring to automated remediation

Most cloud security tooling stops at the alert — a human still has to see the notification, understand the risk, and go fix it. This project asks a narrower question: for a well-known class of misconfigurations, why wait for a human at all? The result is a 24/7 Python auditor that scans AWS infrastructure every 60 seconds and, for the riskiest findings, remediates them itself before paging anyone.

60s

Continuous AWS audit interval

4

Automated risk categories

0-touch

Remediation for exposed firewall rules

24/7

Always-on, cron-driven auditing

01

Why self-healing, not just alerting

A firewall rule that opens SSH (port 22) to the entire internet (0.0.0.0/0) is one of the most common — and most dangerous — cloud misconfigurations. In a purely reactive setup, that window stays open for however long it takes a person to see the alert and act on it. This project’s goal was to close that window automatically: detect the exposure, remove it, and notify the team that it happened — in that order.

02

Four categories of automated defense

Auto-remediation

If a firewall port like SSH 22 is exposed to the open internet, the script deletes the insecure rule immediately — no approval step, no delay.

Identity guard

Flags IAM users without MFA enabled, and access keys older than 90 days — the two most common paths to a compromised AWS account.

Data protection

Identifies unencrypted EBS volumes and publicly accessible S3 buckets before they turn into a data-exposure incident.

Cost control

Monitors running instances to catch “shadow IT” — resources spun up outside normal process — before they turn into budget overruns.

03

How it's built

The auditor itself is a Python script using Boto3(AWS’s SDK) to query account state — firewall rules, IAM users, storage buckets, running instances — on a fixed loop, scheduled with Linux Cron every 60 seconds. The underlying AWS resources it audits are themselves provisioned with Terraform, so the whole environment — the thing being watched and the infrastructure it runs on — is defined as code. When the auditor finds and fixes something, it pushes an instant notification through AWS SNS, so remediation is never silent.

04

Built in two passes

The first working version only covered infrastructure hygiene: unencrypted EBS volumes, missing resource-ownership tags, and insecure firewall rules. Identity was the deliberate next step — after shipping phase one, IAM exposure (users without MFA, stale access keys) was the clearest remaining gap, so it became the focus of the second pass, alongside cost-control checks for unmanaged running instances.

05

What it taught

Security is not a one-time setup — it’s a continuous loop of detection, alerting, and remediation. Automating the routine, well-understood parts of that loop is what frees a security team to spend its attention on the threats that actually need human judgment.