Preventing Open Security Groups: Claude Code and 0.0.0.0/0

Correction note (2026-10-13). An earlier version of this article told readers to export the audit log for a compliance review. ssg has no audit export. The local audit log is the SQLite audit_log table in ~/.sigmashake/rules.db, and ssg audit search queries it. Rows are not signed. The per-call timings and latency figures that appeared in this article had no benchmark behind them and were removed. Sample output in this article is illustrative, not a recorded run. The only published measurement is p50 0.39 ms and p99 1.91 ms (sigmashake-bench daemon-socket run, 2026-06-10, ssg 0.29.156, n=500). Sub-2 ms evaluation is a design target. The local audit log records decisions on a best-effort basis: entries can be dropped if the daemon queue is full, and old rows are deleted at the row caps. It is local, not append-only, and not immutable.

Describe the task

You ask Claude Code to open a port for your API server during a quick test. The agent creates a security group rule using the catch-all CIDR:

aws ec2 authorize-security-group-ingress \
  --group-id sg-xxx \
  --protocol tcp \
  --port 443 \
  --cidr 0.0.0.0/0

0.0.0.0/0 means the entire public internet can reach this port. For a test that becomes a forgotten server, this is an open door. SSG's no-open-security-group rule blocks the API call before it reaches AWS.

# illustrative sample output
[BLOCK] bash › aws ec2 authorize-security-group-ingress ... 0.0.0.0/0 — rule: no-open-security-group

Give SSG context

SSG evaluates Claude Code's shell commands via the claude_code_hook integration (MCP Server or PreToolUse hook).

Required context

rule no-open-security-group {
  enabled true
  priority 95
  severity error
  DENY execution
  IF command CONTAINS "authorize-security-group-ingress"
  AND command CONTAINS "0.0.0.0/0"
  MESSAGE "Public ingress (0.0.0.0/0) is blocked. Specify a VPC CIDR or specific IP range."
}

Optional context

Also block IPv6 open ingress and Terraform resource configurations:

rule no-open-sg-ipv6 {
  enabled true
  priority 95
  severity error
  DENY execution
  IF command CONTAINS "authorize-security-group-ingress"
  AND command CONTAINS "::/0"
  MESSAGE "IPv6 public ingress (::/0) is blocked."
}

rule no-open-sg-terraform {
  enabled true
  priority 90
  severity error
  DENY write
  IF path ENDS_WITH ".tf"
  AND content CONTAINS "cidr_blocks"
  AND content CONTAINS "0.0.0.0/0"
  MESSAGE "Terraform security group rule with 0.0.0.0/0 ingress is blocked."
}

What SSG does

SSG intercepts the aws CLI call before any network request is made to the AWS API. The security group is never modified — the block fires at the tool-call layer.

Ask the agent to use a scoped CIDR

0.0.0.0/0 ingress was blocked. Use your VPC CIDR block or your office IP range instead:
  --cidr 10.0.0.0/8        # VPC-only access
  --cidr 203.0.113.0/24    # your office IP range

Extend to other cloud providers

The same pattern applies to GCP and Azure CLI commands. Create equivalent rules for gcloud compute firewall-rules create and az network nsg rule create:

rule no-gcp-open-firewall {
  enabled true
  priority 95
  severity error
  DENY execution
  IF command CONTAINS "gcloud compute firewall-rules"
  AND command CONTAINS "0.0.0.0/0"
  MESSAGE "GCP firewall rule with public ingress (0.0.0.0/0) is blocked."
}

Audit all security group modifications

Blocked security group changes are logged on a best-effort basis with the full CLI command and arguments. Query the local audit log with ssg audit search for your cloud security review.

How your rule handles multi-line commands

Bash multi-line commands with \ line continuations are joined before SSG evaluates them. The full logical command is the unit of evaluation, so 0.0.0.0/0 appearing on a continuation line is still caught.

Check the rule against your NAT gateway config

Load balancers and public-facing services sometimes legitimately need 0.0.0.0/0 ingress on specific ports. Create an ALLOW exemption rule at higher priority for those known-safe patterns, or use ASK mode to route them to human approval.