Spring boot security

Spring Boot Security Audit

Audit Spring Boot applications (Java and Kotlin, 2.7+ and 3.x).

When this skill applies

• Reviewing Spring Security configuration classes
• Auditing JPA repository methods and queries
• Reviewing controller-level authorization annotations
• Checking actuator endpoint exposure
• Reviewing application.yml / application.properties for secrets

Workflow

Follow ../_shared/audit-workflow.md.

Phase 1: Stack detection

grep -E 'spring-boot-starter' pom.xml build.gradle build.gradle.kts 2>/dev/null
grep -E 'org.springframework' pom.xml 2>/dev/null | head

Phase 2: Inventory

# Security configuration
grep -rn 'SecurityFilterChain\|WebSecurityConfigurerAdapter\|EnableWebSecurity\|EnableMethodSecurity' src/ --include='*.java' --include='*.kt'

# Controllers
grep -rn '@RestController\|@Controller\|@RequestMapping\|@GetMapping\|@PostMapping' src/ --include='*.java' --include='*.kt' | head

# Authorization annotations
grep -rn '@PreAuthorize\|@PostAuthorize\|@Secured\|@RolesAllowed' src/ --include='*.java' --include='*.kt'

# Custom queries
grep -rn '@Query\|@NativeQuery\|createNativeQuery\|createQuery' src/ --include='*.java' --include='*.kt'

# Config files
ls src/main/resources/application*.yml src/main/resources/application*.properties 2>/dev/null

Phase 3: Detection — the checks

Spring Security configuration

Modern Spring Security 6 uses SecurityFilterChain bean. Older used WebSecurityConfigurerAdapter (removed in 6).

• SPR-SC-1 A SecurityFilterChain bean explicitly configured. Don't rely on Spring defaults (they permit-all in older versions).
• SPR-SC-2 Default deny: routes not matched fall through to .anyRequest().authenticated() or .denyAll().
• SPR-SC-3 Public endpoints explicitly allowlisted; everything else requires auth.

@Bean
public SecurityFilterChain filterChain(HttpSecurity http) throws Exception {
    return http
        .authorizeHttpRequests(auth -> auth
            .requestMatchers("/login", "/signup", "/health").permitAll()
            .requestMatchers("/admin/**").hasRole("ADMIN")
            .anyRequest().authenticated())
        .csrf(csrf -> csrf
            .csrfTokenRepository(CookieCsrfTokenRepository.withHttpOnlyFalse()))
        .headers(headers -> headers
            .contentSecurityPolicy(csp -> csp.policyDirectives("default-src 'self'")))
        .build();
}

Authentication

• SPR-AUTH-1 Password encoder is BCrypt, Argon2, or Pbkdf2 — not NoOp.
• SPR-AUTH-2 UserDetailsService returns null-safe results; doesn't leak existence via timing or different error messages.
• SPR-AUTH-3 JWT validation: see saas-security-pack/saas-code-security-review/references/jwt-validation.md. Spring Security OAuth2 Resource Server is the well-trodden path.
• SPR-AUTH-4 Session fixation protection enabled (default in Spring Security; verify not disabled).

Authorization

• SPR-AZ-1 @EnableMethodSecurity on configuration class to enable @PreAuthorize.
• SPR-AZ-2 Service methods that mutate user data have @PreAuthorize("hasRole('USER') and #userId == authentication.principal.id").
• SPR-AZ-3 Controllers use @PreAuthorize OR url-based config — not both inconsistently.
• SPR-AZ-4 @PreFilter / @PostFilter on collection returns to enforce per-element authz.

CSRF

• SPR-CSRF-1 CSRF enabled by default. If disabled (.csrf(csrf -> csrf.disable())), endpoints must be stateless (token auth, no cookie sessions).
• SPR-CSRF-2 REST APIs using JWT in headers can disable CSRF. Cookie-based REST APIs cannot.
• SPR-CSRF-3 CookieCsrfTokenRepository.withHttpOnlyFalse() — the CSRF cookie must be JS-readable for SPAs to send the header; this is correct, not a finding.

CORS

• SPR-COR-1 CORS configured via CorsConfigurationSource bean with specific origins, methods, headers.
• SPR-COR-2 setAllowCredentials(true) only with specific origins.

SQL injection (JPA, JdbcTemplate)

• SPR-SQL-1 @Query with ?1 or named parameters :userId is parameterized.
• SPR-SQL-2 String concatenation in JPQL/native queries is injection:

  // BAD
  @Query(value = "SELECT * FROM users WHERE name = '" + name + "'", nativeQuery = true)
  
  // GOOD
  @Query(value = "SELECT * FROM users WHERE name = :name", nativeQuery = true)
  User findByName(@Param("name") String name);
  

• SPR-SQL-3 JdbcTemplate.queryForObject(sql, ...) uses placeholders; not String.format.
• SPR-SQL-4 Criteria API and Specification queries safe; dynamic identifiers need allowlist.

Jackson deserialization (Spring4Shell-class)

• SPR-JKS-1 Don't deserialize untrusted JSON into polymorphic types (@JsonTypeInfo with default typing). CVE-2017-7525, Spring4Shell (CVE-2022-22965) class.
• SPR-JKS-2 Spring Boot 2.7+ / 3.x patched against the original Spring4Shell vector, but custom Binder configurations may reintroduce — audit any custom WebDataBinder config.
• SPR-JKS-3 @RestController methods accepting Object or generic types are dangerous; use specific DTOs.

Mass assignment via @ModelAttribute

• SPR-MA-1 Controller methods accepting @ModelAttribute User user bind every field. Use DTOs separate from entities:

  @PostMapping("/users")
  public User create(@RequestBody @Valid CreateUserDto dto) {
      // build entity from DTO, set role server-side
  }
  

• SPR-MA-2 WebDataBinder setAllowedFields(...) configured if using @ModelAttribute on entities.

Actuator endpoints

Spring Boot Actuator exposes runtime info. Production exposure can leak sensitive data.

• SPR-ACT-1 management.endpoints.web.exposure.include lists only safe endpoints (health, info). NOT * in production.
• SPR-ACT-2 Sensitive endpoints (heapdump, env, configprops, loggers, mappings, threaddump) disabled or auth-gated.
• SPR-ACT-3 /actuator/health includes only basic status in production (management.endpoint.health.show-details: when-authorized).
• SPR-ACT-4 Actuator on separate management port not reachable from public internet.

management:
  endpoints:
    web:
      exposure:
        include: health,info
  endpoint:
    health:
      show-details: when-authorized
  server:
    port: 8081  # internal-only port

Configuration / secrets

• SPR-CFG-1 Secrets in application.yml use placeholders pulled from env or Vault:

  spring:
    datasource:
      password: ${DB_PASSWORD}
  

• SPR-CFG-2 No committed application-prod.yml with real secrets.
• SPR-CFG-3 Profiles (application-prod.yml, application-dev.yml) loaded based on SPRING_PROFILES_ACTIVE; production profile sets secure defaults.

File uploads

• SPR-UP-1 spring.servlet.multipart.max-file-size and max-request-size set.
• SPR-UP-2 Content type validated by magic bytes (Apache Tika, etc.).

Headers

• SPR-HDR-1 Spring Security headers defaults reasonable; HSTS, X-Content-Type-Options, X-Frame-Options enabled.
• SPR-HDR-2 CSP configured via .headers(h -> h.contentSecurityPolicy(...)).

Logging

• SPR-LOG-1 logging.level not DEBUG/TRACE in production for security-relevant packages (org.springframework.security).
• SPR-LOG-2 Request body logging filters skip sensitive paths (/login, /api/auth).

Dependencies

• SPR-DEP-1 Spring Boot version on supported line (3.x preferred; 2.7 LTS until end of OSS support).
• SPR-DEP-2 mvn dependency-check:check (OWASP Dependency-Check) or gradle dependencyCheckAnalyze clean.
• SPR-DEP-3 Spring Cloud, Spring Data versions compatible with Spring Boot.

Phase 4: Triage

Critical: actuator * exposed; CSRF disabled with cookie sessions; raw SQL with String concatenation; Spring Boot version with unpatched RCE.

Phase 5: Report

Use ../_shared/findings-schema.md. Prefix IDs with SPR-.