Performing Serverless Function Security Review

Performing Serverless Function Security Review

When to Use

• When auditing serverless applications before production deployment
• When investigating potential data exposure through function environment variables or logs
• When assessing the blast radius of a compromised serverless function execution role
• When compliance reviews require documentation of serverless security controls
• When building secure-by-default templates for serverless deployments

Do not use for container or VM security assessments (use container scanning tools), for API security testing (use DAST tools on the API Gateway layer), or for real-time serverless threat detection (use AWS Lambda Extensions with security agents).

Most Often Missed & How to Confirm

• Resolve effective permissions, not just attached policies: combine the execution role's inline policies, AWS managed policies, and any iam:PassRole. A "scoped" managed policy like AWSLambdaBasicExecutionRole plus one wildcard inline statement is the classic silent over-privilege.
• Secrets hide beyond env vars: scan the deployment package and layers for committed .env, config.json, and hardcoded keys. Env-var scanning alone misses code-embedded credentials.
• Two doors to public exposure: Function URL AuthType: NONE AND a resource-based policy with Principal: * on lambda:InvokeFunction. Check both, plus API Gateway authorizers set to NONE.
• Deprecated runtimes (python3.7, nodejs14.x, dotnetcore3.1) no longer receive security patches - flag them even when the code looks clean.
• Event injection: untrusted fields from S3/SQS/SNS/EventBridge events used in shell/SQL/eval. Grep for os.system, subprocess, eval(, string-built queries, pickle.loads, yaml.load.

How to confirm a hit (avoid false negatives): for public access, invoke the Function URL unauthenticated and confirm it actually executes (200 + function output), not merely that AuthType: NONE. For an over-broad role, use aws iam simulate-principal-policy to prove the action is allowed. Don't conclude negative until: the execution role's effective permissions are resolved, the package and layers are scanned for secrets, all invocation paths (Function URL + resource policy + API Gateway) are checked, and CloudWatch logs are reviewed for leaked secrets.

Prerequisites

• AWS CLI, Azure CLI, and gcloud CLI configured with appropriate permissions
• Access to read function configurations, policies, and execution roles
• Prowler or Checkov for automated serverless security scanning
• SAM CLI or Serverless Framework for local function analysis
• CloudTrail, Azure Monitor, or Cloud Audit Logs enabled for function invocation monitoring

Workflow

Step 1: Enumerate All Serverless Functions and Configurations

List all functions across cloud providers with their runtime, memory, timeout, and network settings.

# AWS Lambda: List all functions with key security attributes
aws lambda list-functions \
  --query 'Functions[*].[FunctionName,Runtime,MemorySize,Timeout,Role,VpcConfig.VpcId,Layers[*].Arn]' \
  --output table

# Check for functions using deprecated runtimes
aws lambda list-functions \
  --query 'Functions[?Runtime==`python3.7` || Runtime==`nodejs14.x` || Runtime==`dotnetcore3.1`].[FunctionName,Runtime]' \
  --output table

# Azure Functions: List all function apps
az functionapp list \
  --query "[].{Name:name, Runtime:siteConfig.linuxFxVersion, ResourceGroup:resourceGroup, HttpsOnly:httpsOnly}" \
  -o table

# GCP Cloud Functions: List all functions
gcloud functions list \
  --format="table(name, runtime, status, httpsTrigger.url, serviceAccountEmail, vpcConnector)"

Step 2: Audit Execution Role Permissions

Review IAM roles attached to functions for overly permissive policies.

# AWS: Check each Lambda function's execution role
for func in $(aws lambda list-functions --query 'Functions[*].FunctionName' --output text); do
  role_arn=$(aws lambda get-function-configuration --function-name "$func" --query 'Role' --output text)
  role_name=$(echo "$role_arn" | awk -F'/' '{print $NF}')
  echo "=== $func -> $role_name ==="

  # List attached policies
  aws iam list-attached-role-policies --role-name "$role_name" \
    --query 'AttachedPolicies[*].[PolicyName,PolicyArn]' --output table

  # Check for wildcard actions
  for policy_arn in $(aws iam list-attached-role-policies --role-name "$role_name" --query 'AttachedPolicies[*].PolicyArn' --output text); do
    version=$(aws iam get-policy --policy-arn "$policy_arn" --query 'Policy.DefaultVersionId' --output text)
    aws iam get-policy-version --policy-arn "$policy_arn" --version-id "$version" \
      --query 'PolicyVersion.Document' --output json | python3 -c "
import json, sys
doc = json.load(sys.stdin)
for stmt in doc.get('Statement', []):
    actions = stmt.get('Action', [])
    if isinstance(actions, str): actions = [actions]
    resources = stmt.get('Resource', [])
    if isinstance(resources, str): resources = [resources]
    if '*' in actions or any(a.endswith(':*') for a in actions):
        print(f'  WARNING: {stmt[\"Effect\"]} {actions} on {resources}')
" 2>/dev/null
  done
done

Step 3: Check Environment Variables for Secrets

Scan function environment variables for hardcoded credentials, API keys, and database connection strings.

# AWS Lambda: Extract environment variables
for func in $(aws lambda list-functions --query 'Functions[*].FunctionName' --output text); do
  envvars=$(aws lambda get-function-configuration --function-name "$func" \
    --query 'Environment.Variables' --output json 2>/dev/null)
  if [ "$envvars" != "null" ] && [ -n "$envvars" ]; then
    echo "=== $func ==="
    echo "$envvars" | python3 -c "
import json, sys, re
vars = json.load(sys.stdin)
sensitive_patterns = [
    r'(?i)(password|secret|key|token|credential|api.?key)',
    r'(?i)(aws.?access|aws.?secret)',
    r'(?i)(database.?url|connection.?string|db.?pass)',
    r'AKIA[0-9A-Z]{16}'
]
for key, value in vars.items():
    for pattern in sensitive_patterns:
        if re.search(pattern, key) or re.search(pattern, str(value)):
            masked = value[:4] + '****' + value[-4:] if len(value) > 8 else '****'
            print(f'  SENSITIVE: {key} = {masked}')
            break
"
  fi
done

# Azure Functions: Check app settings
for app in $(az functionapp list --query "[].name" -o tsv); do
  rg=$(az functionapp show --name "$app" --query "resourceGroup" -o tsv)
  echo "=== $app ==="
  az functionapp config appsettings list \
    --name "$app" --resource-group "$rg" \
    --query "[?contains(name,'KEY') || contains(name,'SECRET') || contains(name,'PASSWORD')].{Name:name}" \
    -o table 2>/dev/null
done

Step 4: Review Function Triggers and Access Controls

Verify that function triggers have appropriate authentication and authorization.

# AWS: Check for unauthenticated Lambda function URLs
aws lambda list-function-url-configs \
  --function-name FUNCTION_NAME \
  --query 'FunctionUrlConfigs[*].[FunctionUrl,AuthType,Cors]' --output table

# Check for resource-based policies allowing public invocation
for func in $(aws lambda list-functions --query 'Functions[*].FunctionName' --output text); do
  policy=$(aws lambda get-policy --function-name "$func" --query 'Policy' --output text 2>/dev/null)
  if [ -n "$policy" ]; then
    echo "$policy" | python3 -c "
import json, sys
doc = json.loads(sys.stdin.read())
for stmt in doc.get('Statement', []):
    principal = stmt.get('Principal', {})
    if principal == '*' or principal == {'AWS': '*'}:
        print(f'WARNING: $func has public invoke policy: {stmt.get(\"Sid\", \"unnamed\")}')" 2>/dev/null
  fi
done

# GCP: Check for unauthenticated Cloud Functions
gcloud functions list --format=json | python3 -c "
import json, sys
functions = json.load(sys.stdin)
for func in functions:
    name = func.get('name', '').split('/')[-1]
    trigger = func.get('httpsTrigger', {})
    if trigger and func.get('ingressSettings') == 'ALLOW_ALL':
        print(f'WARNING: {name} allows all ingress traffic')
"

Step 5: Analyze Function Code for Security Vulnerabilities

Review function code for common serverless security issues.

# Download Lambda function code for review
aws lambda get-function --function-name FUNCTION_NAME \
  --query 'Code.Location' --output text | xargs curl -o function.zip
unzip function.zip -d function-code/

# Scan with Bandit (Python) or ESLint security plugin (Node.js)
# Python functions
pip install bandit
bandit -r function-code/ -f json -o bandit-results.json

# Node.js functions
npm install -g eslint @microsoft/eslint-plugin-sdl
eslint --ext .js function-code/

# Check for common serverless vulnerabilities:
# 1. SQL injection in database queries
# 2. Command injection via os.system or subprocess
# 3. Insecure deserialization
# 4. Event data injection (untrusted event parameters)
# 5. Excessive function permissions
grep -rn "os.system\|subprocess\|eval(\|exec(" function-code/ || echo "No obvious injection patterns"
grep -rn "pickle.loads\|yaml.load\b" function-code/ || echo "No deserialization risks"

Step 6: Run Automated Serverless Security Scanning

Execute Checkov and Prowler for automated compliance checks on serverless resources.

# Checkov scan for serverless frameworks
checkov -d ./serverless-project/ \
  --framework serverless \
  --output json > checkov-serverless.json

# Prowler Lambda-specific checks
prowler aws \
  --checks lambda_function_no_secrets_in_variables \
           lambda_function_url_auth_type \
           lambda_function_using_supported_runtimes \
           lambda_function_not_publicly_accessible \
  -M json-ocsf \
  -o ./prowler-lambda/

Key Concepts

Tools & Systems

• AWS Lambda: Primary serverless compute platform with execution roles, layers, and resource policies
• Checkov: Static analysis tool for infrastructure-as-code with serverless-specific security policies
• Prowler: Cloud security tool with Lambda-specific checks for permissions, public access, and runtime versions
• Bandit: Python static analysis tool for detecting security issues in function source code
• OWASP Serverless Top 10: Security risk framework specific to serverless architectures

Common Scenarios

Scenario: Lambda Function with Admin Role Leaking Secrets via Environment Variables

Context: A security review discovers a Lambda function with AdministratorAccess execution role and database credentials stored in plaintext environment variables visible in CloudWatch logs.

Approach:

1. Enumerate the function's execution role and discover AdministratorAccess managed policy
2. Check environment variables and find DB_PASSWORD, API_KEY, and STRIPE_SECRET_KEY in plaintext
3. Review CloudWatch logs and find credentials printed in debug log statements
4. Create a scoped IAM policy granting only the specific DynamoDB and S3 actions needed
5. Migrate secrets to AWS Secrets Manager and update function to retrieve at runtime
6. Remove debug logging that outputs sensitive data
7. Rotate all exposed credentials and enable Lambda function encryption with KMS

Pitfalls: Changing a function's execution role can break it if the new role is too restrictive. Test in a staging environment first. Environment variable changes trigger a new function version, so ensure aliases and triggers are updated. Secrets Manager calls add latency; cache secrets within the execution context to avoid per-invocation lookups.

Output Format

Serverless Function Security Review
=======================================
Account: 123456789012
Functions Reviewed: 34
Review Date: 2026-02-23

CRITICAL FINDINGS:
[SRVL-001] Overly Permissive Execution Role
  Function: payment-processor
  Role: AdministratorAccess (full AWS access)
  Required Permissions: DynamoDB:PutItem, S3:GetObject (2 actions)
  Remediation: Create scoped policy with only required permissions

[SRVL-002] Secrets in Environment Variables
  Function: payment-processor
  Variables: DB_PASSWORD, STRIPE_SECRET_KEY, API_KEY
  Risk: Visible in console, API, and CloudWatch logs
  Remediation: Migrate to Secrets Manager, remove from env vars

SUMMARY:
  Functions with admin roles:           3 / 34
  Functions with secrets in env vars:   8 / 34
  Functions with deprecated runtimes:   5 / 34
  Functions with public access:         2 / 34
  Functions without VPC:               28 / 34
  Functions with wildcard permissions: 12 / 34