tools/iamroot-fleet-scan.sh + docs/DETECTION_PLAYBOOK.md

iamroot-fleet-scan.sh — bash wrapper that scp's the iamroot binary
to a host list, ssh-runs --scan --json on each, aggregates results
into a single JSON document. Supports:
- hosts list from file or stdin
- user@host:port syntax
- parallel xargs execution (default -P 4)
- ssh key / extra ssh opts pass-through
- --no-sudo for hosts where root isn't required
- --summary-only to suppress per-host detail
- --no-cleanup to leave the binary on disk

Critical fix during smoke-test: iamroot's exit codes are SEMANTIC
(0=OK, 2=VULNERABLE, 4=PRECOND_FAIL, 5=EXPLOIT_OK). The wrapper
must NOT treat nonzero exit as a transport failure; success is
defined by 'stdout contains valid JSON', failure by 'stdout empty'.

Verified end-to-end on kctf-mgr → kctf-fuzz:
  fleet-scan reports ok=1, failed=0,
  summary.vulnerable groups by CVE: copy_fail_gcm, dirty_frag_esp×2,
  entrybleed. Per-host detail included.

docs/DETECTION_PLAYBOOK.md — operational integration guide:
- Lifecycle diagram (inventory → scan → fleet scan → deploy/mitigate/upgrade → monitor)
- Recipes by team size: single host, small fleet, large fleet
- SIEM integration patterns: Splunk, Elastic, Sigma
- Auditd-event lookup commands per module key
- VULNERABLE decision tree (patch vs mitigate vs compensate)
- Mitigation revert procedures + side-effect table
- False-positive tuning table per rule key
- Pre-patch quarantine pattern
- Maintenance contract / module-shipping SLA
This commit is contained in:
2026-05-16 20:29:48 -04:00
parent 3eeee01f06
commit ea5d021f0c
3 changed files with 507 additions and 0 deletions
+302
View File
@@ -0,0 +1,302 @@
# IAMROOT detection playbook
Operational guide for blue teams using IAMROOT defensively. Pairs
with `docs/DEFENDERS.md` (the "what" reference) — this is the "how to
make it part of your daily ops" guide.
## The lifecycle
```
┌─────────────┐
│ inventory │ ← iamroot --list (what's bundled?)
└──────┬──────┘
┌─────────────┐
│ scan │ ← iamroot --scan --json (what am I vulnerable to?)
└──────┬──────┘
┌─────────────┐
│ fleet scan │ ← iamroot-fleet-scan.sh hosts.txt
└──────┬──────┘
┌────────────┼────────────┐
▼ ▼ ▼
┌────────┐ ┌─────────┐ ┌──────────┐
│ deploy │ │ mitigate│ │ upgrade │ ← three responses
│ rules │ │ (pre-fix│ │ (kernel │
│(SIEM) │ │ stopgap)│ │ patch) │
└────┬───┘ └─────┬───┘ └─────┬────┘
└────────────┼────────────┘
┌─────────────┐
│ monitor │ ← ausearch -k iamroot-* / SIEM alerts
└─────────────┘
```
## Recipes by team size
### Single host (workstation / single server)
```bash
# Daily/weekly hygiene check
sudo iamroot --scan
# If anything's VULNERABLE, deploy detections + apply mitigation
sudo iamroot --detect-rules --format=auditd | sudo tee /etc/audit/rules.d/99-iamroot.rules
sudo augenrules --load
sudo iamroot --mitigate copy_fail # or whichever module fired
```
### Small fleet (~10-100 hosts, SSH-reachable)
Use `tools/iamroot-fleet-scan.sh`:
```bash
# Hosts list — one per line; user@host:port supported
cat > hosts.txt <<EOF
prod-web-01
prod-web-02
deploy@bastion-01
ops@db-01:2222
EOF
# Scan; binary scp'd, run, cleaned up. Output is one JSON doc.
./iamroot-fleet-scan.sh \
--binary ./iamroot \
--ssh-key ~/.ssh/ops_key \
--parallel 8 \
hosts.txt > fleet-scan-$(date +%F).json
# Show me hosts with any VULNERABLE finding
jq '.hosts[] | select(.scan.modules | map(.result == "VULNERABLE") | any) | .host' \
fleet-scan-*.json
# Show summary across the fleet
jq '.summary' fleet-scan-*.json
```
Output shape:
```json
{
"generated_at": "2026-05-16T22:00:00Z",
"n_hosts": 4,
"summary": {
"ok": 4,
"failed": 0,
"vulnerable": [
{ "cve": "CVE-2024-1086", "name": "nf_tables", "count": 2 },
{ "cve": "CVE-2023-0458", "name": "entrybleed", "count": 4 }
]
},
"hosts": [...]
}
```
### Larger fleet (>100 hosts)
`iamroot-fleet-scan.sh` is intentionally simple (parallel ssh). For
fleets too large for SSH-fan-out, wrap it in your config-management
tool of choice:
- **Ansible**: ship the binary via `copy:`, run via `command:`, parse
JSON with `jq` in a follow-on task
- **SaltStack**: `cmd.run` returning JSON; `salt-call --return` to your
SIEM
- **Fabric / Mitogen**: same shape, just Python-side
Sample Ansible task:
```yaml
- name: scan with iamroot
copy:
src: iamroot
dest: /tmp/iamroot
mode: '0755'
- name: run --scan --json
command: /tmp/iamroot --scan --json --no-color
register: scan
changed_when: false
failed_when: false # iamroot exit codes are semantic, not errors
- name: collect
set_fact:
iamroot_scan: "{{ scan.stdout | from_json }}"
- name: cleanup
file:
path: /tmp/iamroot
state: absent
```
## SIEM integration patterns
### Splunk
```
# splunk input config (inputs.conf)
[script:///opt/iamroot/iamroot-cron-scan.sh]
interval = 86400
source = iamroot
sourcetype = iamroot:scan
```
`iamroot-cron-scan.sh`:
```bash
#!/bin/bash
/usr/local/bin/iamroot --scan --json --no-color
```
Search the indexed events:
```spl
index=iamroot sourcetype="iamroot:scan" modules{}.result=VULNERABLE
| stats count by host modules{}.cve
```
### Elastic / OpenSearch
Filebeat module reading the per-host scan JSON files (one per day),
indexed into an `iamroot-*` index pattern. Standard Kibana
visualization on `modules.cve` over time tracks vulnerability lifecycle.
### Sigma → your platform
```bash
# Ship Sigma rules into your platform
iamroot --detect-rules --format=sigma > /etc/sigma/iamroot.yml
# Convert to your target (Sentinel, Elastic, etc.) via sigmac
sigmac -t elastic /etc/sigma/iamroot.yml
```
## Day-to-day operational shape
### What "good" looks like in the SIEM
- Daily `iamroot --scan --json` from every host indexed
- Trend dashboard: count of VULNERABLE results by CVE over time
- Goal: every VULNERABLE → OK transition within SLA (e.g., 14 days for
patched-mainline bugs, 24h for actively-exploited)
- Alert on: any host with a result not seen yesterday (could indicate
a config drift, a new install, or a disabled mitigation)
### Auditd events from the embedded rules
After deploying `iamroot --detect-rules --format=auditd`:
```bash
# By module key
sudo ausearch -k iamroot-copy-fail -ts today
sudo ausearch -k iamroot-dirty-pipe -ts today
sudo ausearch -k iamroot-pwnkit -ts today
sudo ausearch -k iamroot-nf-tables-userns -ts today
sudo ausearch -k iamroot-overlayfs -ts today
# Anything iamroot-tagged in the last hour
sudo ausearch -k 'iamroot-*' -ts recent
# Forward to syslog (rsyslog example)
# /etc/rsyslog.d/iamroot.conf:
:msg, contains, "iamroot-" @@your-siem.example.com:514
```
### When a VULNERABLE result fires
Decision tree:
```
A scan reports VULNERABLE for module X
├── Q: Can I patch the underlying kernel / package?
│ ├── YES → schedule patch window. In the meantime:
│ │ iamroot --mitigate X (if supported)
│ │ Verify auditd rule for X is loaded.
│ │ Monitor for the rule key.
│ └── NO (legacy LTS, embedded device, prod freeze) →
│ iamroot --mitigate X (essential)
│ Compensating control: tighten LSM (SELinux/AppArmor)
│ Document in risk register
└── Q: Was this VULNERABLE before? When?
├── First time → config drift; investigate why detection now
│ produces this result
└── Persistent → mitigation isn't applied OR is being reverted
by config management; fix the config baseline
```
### Mitigation reverts
Mitigations can break legitimate functionality:
| Mitigation | Side effect |
|---|---|
| `copy_fail` blacklist algif_aead | strongSwan / IPsec breaks |
| `copy_fail` blacklist esp4/esp6 | IPsec breaks |
| `copy_fail` blacklist rxrpc | AFS / kAFS clients break |
| `copy_fail` AppArmor restrict userns=1 | bubblewrap, podman rootless break |
If you applied a mitigation and now need to revert (e.g., the kernel
patch has rolled out fleet-wide):
```bash
sudo iamroot --cleanup copy_fail
# OR manually:
sudo rm /etc/modprobe.d/dirtyfail-mitigations.conf
sudo rm /etc/sysctl.d/99-dirtyfail-mitigations.conf
# Reload affected modules / sysctls per your distro
```
## Common false positives + tuning
| Rule key | False positive | Fix |
|---|---|---|
| `iamroot-copy-fail-afalg` | strongSwan, libcrypto using kernel crypto | `-F auid=` exclude service account UIDs |
| `iamroot-dirty-pipe-splice` | nginx, HAProxy, kTLS | `-F gid!=33 -F gid!=99` exclude web service accounts |
| `iamroot-pwnkit-execve` | gnome-software, polkit's own re-exec | Correlate by parent process; pkexec via gnome dbus is benign |
| `iamroot-nf-tables-userns` | docker rootless, podman, snap confined apps | Whitelist known userns-using service GIDs |
| `iamroot-overlayfs` | docker / containerd mounting overlayfs as root | The rule is intended for unprivileged-userns overlayfs mounts; add `-F auid>=1000` |
## Pre-patch quarantine pattern
If a CVE is in active exploitation and you can't patch immediately:
```bash
# Stage 1: detect
sudo iamroot --scan --json | jq '.modules[] | select(.cve == "CVE-XXXX")'
# Stage 2: mitigate (where supported)
sudo iamroot --mitigate <module>
# Stage 3: monitor — auditd rules already deployed
sudo ausearch -k 'iamroot-*' -ts today | grep <module>
# Stage 4: contain — temporarily restrict the trigger surface
# e.g., for nf_tables CVE-2024-1086:
echo 0 | sudo tee /proc/sys/kernel/unprivileged_userns_clone
# OR
sudo sysctl -w kernel.apparmor_restrict_unprivileged_userns=1
# Stage 5: alert
# When auditd or sigma rule fires, page on-call
```
## Maintenance contract
When IAMROOT ships a new module:
1. CI test passes on at least one vulnerable + patched kernel pair
2. Detection rules ship alongside (auditd + sigma minimum)
3. CVES.md row added with patch status
4. NOTICE.md credits original researcher
5. ROADMAP.md updated
Treat these as the SLA for any blue-team-facing deliverable.
## When you find a new false positive
File an issue at https://github.com/KaraZajac/IAMROOT/issues with:
- The exact ausearch line that fired
- The legitimate process that produced it
- Distro / kernel version
Most false-positive fixes are a `-F` filter on the embedded rule —
small, mergeable.
Executable
BIN
View File
Binary file not shown.
+205
View File
@@ -0,0 +1,205 @@
#!/usr/bin/env bash
# iamroot-fleet-scan — scan a host list with iamroot, aggregate results
#
# Usage:
# iamroot-fleet-scan.sh [OPTIONS] hosts.txt
# iamroot-fleet-scan.sh [OPTIONS] - # hosts on stdin
# iamroot-fleet-scan.sh [OPTIONS] - # one host per line
#
# Each line in the host list is either:
# - a hostname/IP (uses default ssh user from your config)
# - user@host
# - user@host:port
#
# Output: combined JSON to stdout, one object per host:
# { "generated_at": "...", "summary": {...},
# "hosts": [ { "host": "...", "ok": true,
# "scan": { /* iamroot --scan --json */ } }, ... ] }
#
# Options:
# --binary <path> path to iamroot binary (default: ./iamroot)
# --ssh-key <path> ssh key file (passed to scp and ssh)
# --ssh-opts "..." extra ssh options (e.g. "-o ConnectTimeout=5")
# --remote-path <p> where to scp the binary (default: /tmp/iamroot)
# --no-sudo don't prefix the remote command with sudo
# --parallel <N> run N hosts concurrently (default: 4)
# --summary-only skip per-host detail in stdout; print summary only
# --no-cleanup leave the binary behind on each host (default: rm)
# -h | --help this message
#
# Exit code: 0 if every host scanned (regardless of host-level vulns),
# 1 if any host failed to scan.
set -euo pipefail
BINARY="./iamroot"
SSH_KEY=""
SSH_OPTS=""
REMOTE_PATH="/tmp/iamroot"
USE_SUDO=1
PARALLEL=4
SUMMARY_ONLY=0
CLEANUP=1
HOSTFILE=""
usage() { sed -n '2,/^$/p' "$0" | sed 's/^# \?//'; exit "${1:-0}"; }
while [[ $# -gt 0 ]]; do
case "$1" in
--binary) BINARY="$2"; shift 2;;
--ssh-key) SSH_KEY="$2"; shift 2;;
--ssh-opts) SSH_OPTS="$2"; shift 2;;
--remote-path) REMOTE_PATH="$2"; shift 2;;
--no-sudo) USE_SUDO=0; shift;;
--parallel) PARALLEL="$2"; shift 2;;
--summary-only) SUMMARY_ONLY=1; shift;;
--no-cleanup) CLEANUP=0; shift;;
-h|--help) usage 0;;
-) HOSTFILE="/dev/stdin"; shift;;
*) HOSTFILE="$1"; shift;;
esac
done
if [[ -z "$HOSTFILE" ]]; then
echo "error: no host file provided. Use -h for help." >&2
exit 2
fi
if [[ ! -x "$BINARY" ]]; then
echo "error: iamroot binary not found / not executable: $BINARY" >&2
exit 2
fi
if ! command -v jq >/dev/null 2>&1; then
echo "error: jq is required for JSON aggregation" >&2
exit 2
fi
# Build ssh/scp option arrays
SSH_BASE=(-o StrictHostKeyChecking=no -o BatchMode=yes -o ConnectTimeout=10)
[[ -n "$SSH_KEY" ]] && SSH_BASE+=(-i "$SSH_KEY")
[[ -n "$SSH_OPTS" ]] && eval "SSH_BASE+=( $SSH_OPTS )"
scan_one_host() {
local hostspec="$1"
local host port user
if [[ "$hostspec" == *:* ]]; then
port="${hostspec##*:}"
hostspec="${hostspec%:*}"
else
port="22"
fi
if [[ "$hostspec" == *@* ]]; then
user="${hostspec%@*}"
host="${hostspec#*@}"
else
user=""
host="$hostspec"
fi
local target="${user:+${user}@}${host}"
local sudo_prefix=""
[[ "$USE_SUDO" -eq 1 ]] && sudo_prefix="sudo"
# 1. scp the binary
if ! scp "${SSH_BASE[@]}" -P "$port" -q "$BINARY" \
"${target}:${REMOTE_PATH}" 2>/dev/null; then
echo "{\"host\":\"${hostspec}\",\"ok\":false,\"error\":\"scp failed\"}"
return 1
fi
# 2. run --scan --json
# iamroot's exit codes are SEMANTIC (0=OK, 2=VULNERABLE, 4=PRECOND_FAIL, etc.)
# — nonzero is NOT a failure here. Treat ANY stdout JSON as success;
# only ssh-transport-level failures (key denied, network) are real
# failures, and those manifest as empty stdout + nonzero exit.
local scan_out
scan_out=$(ssh "${SSH_BASE[@]}" -p "$port" "$target" \
"$sudo_prefix $REMOTE_PATH --scan --json --no-color" 2>/dev/null || true)
if [[ -z "$scan_out" ]]; then
echo "{\"host\":\"${hostspec}\",\"ok\":false,\"error\":\"ssh run failed (empty output)\"}"
# Still try to cleanup
[[ "$CLEANUP" -eq 1 ]] && ssh "${SSH_BASE[@]}" -p "$port" "$target" \
"rm -f $REMOTE_PATH" 2>/dev/null || true
return 1
fi
# 3. cleanup
if [[ "$CLEANUP" -eq 1 ]]; then
ssh "${SSH_BASE[@]}" -p "$port" "$target" \
"rm -f $REMOTE_PATH" 2>/dev/null || true
fi
# 4. emit one combined JSON object
if ! echo "$scan_out" | jq --arg h "$hostspec" \
'{host: $h, ok: true, scan: .}' 2>/dev/null; then
echo "{\"host\":\"${hostspec}\",\"ok\":false,\"error\":\"invalid JSON from iamroot\"}"
return 1
fi
}
# Read host list (strip comments, blank lines)
mapfile -t HOSTS < <(grep -vE '^\s*(#|$)' "$HOSTFILE")
if [[ ${#HOSTS[@]} -eq 0 ]]; then
echo "error: no hosts to scan" >&2
exit 2
fi
# Optional progress to stderr
echo "[*] scanning ${#HOSTS[@]} host(s), parallel=$PARALLEL" >&2
# Run in parallel with xargs. Each invocation prints one JSON object.
export -f scan_one_host
export BINARY SSH_BASE SSH_KEY REMOTE_PATH USE_SUDO CLEANUP
# bash-export of an array doesn't survive, so re-serialize:
export SSH_BASE_STR="${SSH_BASE[*]}"
# Simpler: collect per-host results sequentially (good enough for small
# fleets); parallel mode uses GNU xargs -P if available.
TMP=$(mktemp)
trap 'rm -f "$TMP"' EXIT
if [[ "$PARALLEL" -gt 1 ]] && command -v xargs >/dev/null 2>&1; then
# -I{} implies -n1; specifying both warns on modern xargs.
printf '%s\n' "${HOSTS[@]}" | xargs -P"$PARALLEL" -I{} \
bash -c 'scan_one_host "$@"' _ {} >> "$TMP"
else
for h in "${HOSTS[@]}"; do
scan_one_host "$h" >> "$TMP" || true
done
fi
# Aggregate. `jq -s` slurps the line-delimited JSON into an array.
TIMESTAMP=$(date -u +%Y-%m-%dT%H:%M:%SZ)
RESULT=$(jq -s --arg ts "$TIMESTAMP" '
. as $hosts
| {
generated_at: $ts,
n_hosts: ($hosts | length),
summary: {
ok: ($hosts | map(select(.ok)) | length),
failed: ($hosts | map(select(.ok | not)) | length),
vulnerable: (
$hosts
| map(select(.ok))
| map(.scan.modules // [])
| flatten
| map(select(.result == "VULNERABLE"))
| group_by(.cve)
| map({cve: .[0].cve, name: .[0].name, count: length})
| sort_by(-.count)
)
},
hosts: $hosts
}
' "$TMP")
if [[ "$SUMMARY_ONLY" -eq 1 ]]; then
echo "$RESULT" | jq 'del(.hosts)'
else
echo "$RESULT"
fi
# Exit nonzero if any host failed
FAILED=$(echo "$RESULT" | jq -r '.summary.failed')
[[ "$FAILED" -eq 0 ]] || exit 1