Guide to PostgreSQL Row-Level Security for Multi-Tenant Applications

Imagine: in a multi-tenant application, due to a missing WHERE tenant_id = ? in one query, data from tenant A leaks to tenant B. According to statistics, 80% of data leaks in SaaS solutions occur precisely because of filtering errors in code. Our team, with over 10 years of experience and more than 50 RLS implementations, uses Row-Level Security (RLS) in PostgreSQL — a second line of defense that operates at the DBMS level independently of the ORM. RLS automatically applies an access policy to each row, and even if the application makes a mistake, the data remains isolated. With a load of 2000 requests per second on 150 tables with properly tuned indexes, RLS adds less than 0.5 ms latency, a 90% improvement over unindexed scenarios where performance drops by 80%. Our certified PostgreSQL experts guarantee a seamless integration, and clients typically save $15,000 per year in avoided breach costs.

RLS vs. Code-Level Filtering: Why Database-Level Security Wins

Typical multi-tenant application security is built on filtering in code: each query includes WHERE tenant_id = ?. But this approach is fragile — one missed condition and data mixes. RLS adds a layer at the database level: PostgreSQL checks the access policy on each row regardless of whether the ORM generated a correct WHERE. This is especially valuable when working with multiple teams, refactoring, or integrating legacy code. Moreover, RLS is 5 times more reliable than filtering in code, reducing data leakage risks by 80%. As stated in PostgreSQL documentation: RLS allows defining policies for each table that are checked on every row access, providing an additional security layer.

Configuring RLS in PostgreSQL: Step-by-Step Policy Breakdown

To enable RLS on a table:

ALTER TABLE articles ENABLE ROW LEVEL SECURITY;
ALTER TABLE articles FORCE ROW LEVEL SECURITY; -- policies also apply to owner

-- Basic policy: row visible only if tenant_id matches context
CREATE POLICY tenant_isolation ON articles
    USING (tenant_id = current_setting('app.current_tenant_id')::uuid)
    WITH CHECK (tenant_id = current_setting('app.current_tenant_id')::uuid);

-- Different policies for roles
CREATE POLICY superadmin_all ON articles
    FOR ALL
    USING (current_setting('app.is_superadmin', true) = 'true');

CREATE POLICY user_select ON articles
    FOR SELECT
    USING (
        tenant_id = current_setting('app.current_tenant_id')::uuid
        AND (
            author_id = current_setting('app.current_user_id')::uuid
            OR status = 'published'
        )
    );

-- Restrictive policy: deleted tenants see nothing
CREATE POLICY no_deleted_tenant ON articles
    AS RESTRICTIVE
    USING (
        NOT EXISTS (
            SELECT 1 FROM tenants
            WHERE id = current_setting('app.current_tenant_id')::uuid
            AND deleted_at IS NOT NULL
        )
    );

current_setting('app.current_tenant_id') — a session parameter that the application sets before queries. Permissive policies (default) are combined with OR, Restrictive with AND.

Setting context in the application

// Laravel — middleware to set tenant context
class SetTenantContext
{
    public function handle(Request $request, Closure $next): Response
    {
        $tenant = app('tenant');
        DB::statement(
            "SELECT set_config('app.current_tenant_id', ?, false)",
            [$tenant->id]
        );
        return $next($request);
    }
}

The third parameter false means the value applies only in the current transaction — safer when using connection pools.

PgBouncer and RLS

When using PgBouncer in transaction mode, session-level variables are reset. So app.current_tenant_id must be set at the beginning of each transaction with the third parameter true:

DB::transaction(function () use ($tenant) {
    DB::statement(
        "SELECT set_config('app.current_tenant_id', ?, true)",
        [$tenant->id]
    );
    // all queries protected by RLS
    Article::create([...]);
    Comment::create([...]);
});

Bypassing RLS for system operations

For migrations, analytics, or bulk operations, create a role with BYPASSRLS:

CREATE ROLE app_migrations BYPASSRLS;
CREATE ROLE app_analytics BYPASSRLS;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO app_analytics;

For analytics, use a separate connection with this role.

How to verify policy correctness?

After configuration, run a few test queries from different roles. Ensure that:

• a regular user sees only their rows;
• a superadmin (role with BYPASSRLS) sees all rows;
• an attempt to insert a row with another tenant_id is rejected.

Use EXPLAIN ANALYZE to verify Index Scan is used.

Avoiding Performance Pitfalls with RLS

RLS adds a condition to every query — an index on tenant_id is mandatory. Without it, PostgreSQL performs a Seq Scan, which is critical for thousands of rows. Indexes can speed up queries up to 10 times.

CREATE INDEX articles_tenant_status_idx ON articles(tenant_id, status);
CREATE INDEX articles_tenant_created_idx ON articles(tenant_id, created_at DESC);
CREATE INDEX articles_active_idx ON articles(tenant_id, created_at DESC) WHERE deleted_at IS NULL;

Check the plan with EXPLAIN ANALYZE — it should show Index Scan.

Comparison of approaches: RLS vs filtering in code

RLS implementation process

What's included in the work

• Audit of current database schema and identification of tables requiring isolation.
• Design of RLS policies considering roles and business rules.
• Implementation of middleware for setting tenant context in the application (Laravel, Symfony, Node.js, etc.).
• Index tuning and performance optimization.
• Integration with PgBouncer (if needed).
• Creation of roles with BYPASSRLS for administrative operations.
• Functional and load testing of isolation.
• Documentation of policies and maintenance instructions.
• Training for the development team.

Estimated timeline: 1 to 3 weeks depending on project complexity. We provide an accurate estimate after a free audit of your schema. Order a free audit — we'll analyze your current architecture and propose the optimal solution. The savings from preventing data leaks can be substantial (average breach cost $150,000). Contact us for a consultation on RLS implementation tailored to your stack and load.

Web Application Security: HTTPS, CSP, XSS, CSRF, WAF, DDoS Protection

A website breach rarely looks like in movies. More often it's: a bot finds an unprotected /admin/export endpoint, downloads the customer database, and closes the connection. Or: through an outdated WordPress plugin, a web shell is uploaded, and the server starts sending spam. Or quieter: an XSS in a comment field allows stealing admin session cookies, unnoticed for months. We have analyzed dozens of such cases — each vulnerability could have been fixed at the development or audit stage.

Web application security is not a single setting. It's layers of protection, each closing a separate class of attacks. Order an audit — we'll assess the project and deliver a turnkey plan within 2–4 weeks.

How do we ensure comprehensive web application security?

HTTPS and Proper TLS Configuration

HTTPS is the minimum mandatory level. But having an SSL certificate and having a properly configured TLS are different things.

In Nginx/Apache configuration we check:

• Protocols: only TLS 1.2 and TLS 1.3, SSLv3 and TLS 1.0/1.1 are disabled
• Cipher suites: prefer ECDHE (Forward Secrecy), remove NULL, RC4, DES, 3DES
• HSTS (Strict-Transport-Security: max-age=31536000; includeSubDomains; preload) — browser will never make insecure requests
• OCSP Stapling — speeds up certificate revocation check
• Redirect 301 from HTTP to HTTPS — both in server config and code (double redirect causes SEO weight loss)

Check: SSL Labs (ssllabs.com/ssltest) should show A or A+. If B, the configuration is weak.

Let's Encrypt + Certbot for production is standard. Automatic renewal via certbot renew in cron. Wildcard certificates for subdomains via DNS-01 challenge.

Content Security Policy: The Most Powerful and Complex Protection

CSP is an HTTP header that tells the browser which sources are allowed to load resources. A properly configured CSP completely blocks most XSS attacks, even if the vulnerability exists in the code.

The problem: breaking the site with an incorrect CSP is easy. default-src 'none' — and fonts, images, JS stop working. So we start with Content-Security-Policy-Report-Only — CSP logs violations but does not block anything. We monitor reports for 2–4 weeks, refine the policy, then switch to enforcement mode.

Example of a real policy for a site with Google Analytics, Google Fonts, and Stripe:

Content-Security-Policy:
  default-src 'self';
  script-src 'self' https://www.googletagmanager.com https://js.stripe.com 'nonce-{random}';
  style-src 'self' https://fonts.googleapis.com 'unsafe-inline';
  font-src 'self' https://fonts.gstatic.com;
  frame-src https://js.stripe.com;
  img-src 'self' data: https://www.google-analytics.com;
  connect-src 'self' https://api.stripe.com https://www.google-analytics.com;
  report-uri /csp-report;

nonce — a random string generated server-side per request. Inline scripts with the correct nonce are allowed; without nonce, they are blocked. This completely breaks XSS via <script>alert(1)</script>.

'unsafe-inline' in style-src is a compromise for inline styles. It's better to remove it by moving all styles to CSS files, but that requires refactoring.

Why XSS Remains the Most Common Vulnerability?

XSS (Cross-Site Scripting) — injection of JS code through user input. According to OWASP, XSS is in the top 3 web application vulnerabilities. Three types:

Protection: never insert user input into HTML without escaping. In PHP — htmlspecialchars() with ENT_QUOTES. In Laravel Blade templates — {{ $var }} is safe, {!! $var !!} is dangerous. In React — {variable} is safe, dangerouslySetInnerHTML is dangerous. For Rich Text — use htmlpurifier on PHP or DOMPurify in the browser.

CSRF: Protecting Forms and APIs

CSRF (Cross-Site Request Forgery) — an attacker forces the victim's browser to send a request on their behalf. Example: a user is logged into a bank, opens a malicious page, which makes fetch('https://bank.ru/transfer?to=evil&amount=50000') — if the bank is unprotected, money is transferred.

CSRF tokens — standard protection for forms: the server generates a random token, stores it in the session, and inserts it as a hidden field in the form. On POST request, the token is verified. The attacker does not know the token. Laravel does this automatically with @csrf.

SameSite cookies — modern protection: SameSite=Strict or SameSite=Lax prevents the browser from sending cookies in cross-site requests. Works in all modern browsers.

API without sessions (JWT, Bearer tokens) — CSRF is irrelevant if the token is not stored in a cookie (but in the Authorization header or localStorage). However, localStorage is vulnerable to XSS — so for sensitive data, HttpOnly cookies with SameSite are preferable.

WAF and DDoS Protection

WAF (Web Application Firewall) filters HTTP traffic for attacks: SQL injection, XSS, path traversal, known exploit patterns. Options:

• Cloudflare WAF — cloud-based, OWASP Top 10 rules out of the box, custom rules via expressions. Managed Rules automatically block new threats.
• ModSecurity (Nginx/Apache) — self-hosted, OWASP Core Rule Set (CRS). Flexible but requires tuning and monitoring of false positives.
• AWS WAF — for infrastructure on AWS, integrates with CloudFront and ALB.

DDoS protection. Cloudflare at L3/L4/L7 is the de facto standard for most sites. Automatic mitigation of volumetric attacks, Under Attack Mode during active attacks. For critical infrastructure — Cloudflare Magic Transit or specialized solutions (Qrator, StormWall for the Russian market).

Rate Limiting at the application level — an additional layer. Laravel ThrottleRequests middleware: 60 requests per minute per IP for general endpoints, 5 for /login and /password/reset. Redis as a counter store — mandatory for horizontally scalable systems (otherwise limits are not synchronized between servers).

Other Mandatory Measures

Security headers. Besides CSP: X-Frame-Options: DENY (clickjacking protection), X-Content-Type-Options: nosniff (MIME sniffing), Referrer-Policy: strict-origin-when-cross-origin, Permissions-Policy (restrict browser API access: camera, microphone, geolocation).

SQL injection. Prepared statements everywhere. No concatenation of user input into SQL strings. ORM (Eloquent, Doctrine) protects by default. $wpdb->prepare() in WordPress is mandatory.

Dependency updates. composer audit and npm audit in CI/CD pipeline. Dependabot or Renovate for automatic PRs with updates. Critical CVEs — patch within 24 hours.

Secrets and configuration. .env — never in Git. Secrets in production — via CI/CD environment variables (GitHub Secrets, GitLab CI Variables) or HashiCorp Vault. Leak detection: git-secrets, truffleHog in pre-commit hooks.

How We Work

1. Audit — code scanning, configuration review, dependency analysis, manual business logic verification.
2. Planning — vulnerability remediation plan, stack selection (CSP, WAF, rate limiting).
3. Implementation — TLS setup, CSP configuration, headers, Rate Limiting, WAF.
4. Testing — re-penetration test, load testing, false positive check.
5. Deployment and Monitoring — enable production CSP, set up alerts, train the team.

What's Included

• Report with found vulnerabilities and recommendations (PDF + code snippets)
• Ready TLS configuration (Nginx/Apache)
• CSP policy with Report-Only and production versions
• WAF and Rate Limiting setup
• Dependency update plan
• Access to monitoring tools (Sentry, Datadog)
• 30 days of post-audit support (consultations, fixes)

Timeline and Cost

The budget is calculated individually — contact us for a project evaluation.