Cloud & Security · SaaS & Startups

API Security for Small SaaS Teams: A Practical 2026 Guide

The OWASP API Top 10 mapped to the controls a five to twenty person SaaS team can actually build, test and maintain.

XDQ Editors7 Nov 2026 · 10 min read
Cardboard Box

For most SaaS products the API is the product, and attackers know it. Salt Security's 2025 State of API Security Report found that 99% of surveyed organisations had API security issues in the previous 12 months, and 95% of API attacks came from authenticated users. This guide maps the OWASP API Security Top 10 to the controls a small team can build without a dedicated security department.

99%
Organisations with API security issues in the past year (Salt, 2025)
Source: Infosecurity Magazine
95%
API attacks that came from authenticated users (Salt, 2025)
Source: Infosecurity Magazine
98%
API attack attempts aimed at external-facing APIs (Salt, 2025)
Source: Infosecurity Magazine
22%
Breaches where credential abuse was the initial vector (DBIR 2025)
Source: Verizon

Why authenticated users are the main API risk

Many teams treat login as the security boundary: once a request carries a valid token, it is trusted. Salt's data shows why that is not enough. If 95% of attacks come from authenticated users, the real question for each endpoint is not whether the caller is logged in but whether this caller may act on this specific object. This is also why API flaws often slip past scanners. A request that reads another tenant's record looks syntactically perfect. Only business rules, written into your code and tests, can tell the difference.

The OWASP API Security Top 10 (2023) for SaaS

OWASP's API list was last refreshed in 2023 and remains the reference most buyers ask about. The table translates each item into what it usually looks like in a SaaS codebase, with the first control worth building. Several of these, notably API1, API3 and API5, are authorisation problems, which is why they rarely show up in automated scans and need tests written around your own business rules.

RiskWhat it looks like in SaaSFirst control to build
API1 Broken Object Level AuthorizationChanging an ID returns another tenant's recordOwnership check on every object lookup
API2 Broken AuthenticationWeak tokens, long-lived keys, no MFA for adminsStandard OAuth or OIDC flows, short token lifetimes
API3 Broken Object Property Level AuthorizationResponse leaks fields, or client can set 'role'Explicit response schemas and allowlisted writable fields
API4 Unrestricted Resource ConsumptionUnlimited exports, SMS or email triggersRate limits, pagination and quotas
API5 Broken Function Level AuthorizationNormal user calls an admin endpointRole checks on every route, deny by default
API6 Unrestricted Access to Sensitive Business FlowsBots abuse sign-up, coupons or bookingsPer-flow limits and abuse monitoring
API7 Server Side Request ForgeryWebhook or URL import fetches internal addressesURL allowlists and egress controls
API8 Security MisconfigurationVerbose errors, open CORS, default settingsHardened gateway and framework config
API9 Improper Inventory ManagementOld v1 endpoints and debug routes still liveSingle API spec as source of truth
API10 Unsafe Consumption of APIsTrusting data from partner or third-party APIsValidate and limit third-party responses

Fixing object-level authorisation properly

BOLA is first on the OWASP API list, and in multi-tenant SaaS it is the flaw most likely to cause a reportable breach. The durable fix is architectural: scope every database query by tenant and user at the data-access layer, so a developer cannot forget the check in a single handler.

Back it with tests. For each resource type, create two tenants in your test suite and assert that tenant A receives a not-found or forbidden response for every one of tenant B's IDs. Random, non-sequential IDs make enumeration harder, but they are not a substitute for the check.

Authentication, tokens and keys

Use a well-supported identity provider or framework rather than writing token handling yourself. Curity's guidance on the OWASP list recommends going beyond simple API keys, using OAuth flows for users and mutual TLS or sender-constrained tokens for machine clients where the risk justifies it.

Keys deserve special care because they leak. GitGuardian's 2025 report found MongoDB credentials were the most common secret leaked in public repositories, at 18.8% of cases. Give customer API keys scopes and expiry dates, show only a prefix after creation, and let customers rotate them without a support ticket.

Rate limits and business-flow protection

Every request costs something: compute, database load, or a paid SMS or email. API4 and API6 are about stopping one caller from consuming far more than intended, whether through a script that exports everything or a bot farming free trials. Apply limits per token and per IP at the gateway, and stricter limits on login, password reset, invitations and any endpoint that sends messages. For business flows, ask what an attacker would gain by automating each step. Limits on resource creation over a time window and a step-up check for unusual volumes usually stop the abuse without hurting real users.

Inventory, versioning and third-party APIs

You cannot protect an endpoint you have forgotten. Salt found that 58% of organisations monitor APIs less than daily and lack confidence in their inventory. Keep an OpenAPI specification in the repository, generate documentation from it, and fail the build when a route exists in code but not in the spec. Set a retirement date for every old version. API10 reverses the usual view: your service is the consumer. Treat responses from payment, CRM or messaging partners as untrusted input. Validate their shape, cap their size, set timeouts and never pass their data straight into queries.

Logging and testing APIs on a small team

A small team can cover most API risk with a short, repeatable routine rather than an expensive platform. The checklist below fits into normal sprint work. The OWASP Logging Cheat Sheet recommends recording authorisation failures and suspicious business logic, such as actions that exceed limits, and those events are the raw material for most of the alerts listed here.

  • Log every authorisation failure with user, tenant, endpoint and object ID
  • Alert on one user hitting many distinct object IDs in a short window
  • Run cross-tenant authorisation tests in CI for every new endpoint
  • Review the API spec diff in every pull request
  • Test webhooks and URL imports against internal address ranges
  • Book an authorised external API test before major enterprise launches

Where to start this quarter

If you only do three things, make them tenant-scoped data access, rate limits on sensitive endpoints, and an accurate API inventory. Together they address the categories most often behind real SaaS incidents. IBM's 2025 study found breaches detected internally cost about USD 900,000 less than those disclosed by an attacker, which is a strong argument for the logging step too. XDQ Labs builds and tests SaaS APIs for teams in India, the US, the UAE and the UK, and the same checklist applies whatever the stack.

Frequently asked questions

What is the most common API vulnerability?

Broken Object Level Authorization (BOLA) is first on the OWASP API Security Top 10 2023. It happens when an endpoint accepts an object ID and returns or changes that object without checking the caller is allowed to access it. In multi-tenant SaaS it can expose one customer's data to another.

Is an API gateway enough for API security?

No. A gateway helps with authentication, rate limits and configuration, but it cannot know whether a user owns a specific record. Object and function level authorisation must be enforced in your application code and data layer, and tested automatically.

How often should a small SaaS team test its APIs?

Automated authorisation and schema tests should run on every pull request. A deeper manual or external test makes sense before major launches, enterprise security reviews or big architecture changes. Logging and alerting give continuous coverage between tests.

What is unsafe consumption of APIs?

It is API10 in the OWASP 2023 list. It describes trusting data from third-party or partner APIs more than user input. If a partner is compromised or returns unexpected data, your service can be attacked through it. Validate, size-limit and time-limit every third-party response.

Sources

  1. OWASP API Security Project: API Security Top 10 2023, OWASP Foundation, 2023
  2. Top 10 API Security Vulnerabilities According to OWASP, Curity, 2024
  3. 99% of Organizations Report API-Related Security Issues, Infosecurity Magazine, 2025
  4. Verizon 2025 Data Breach Investigations Report: news release, Verizon, 2025
  5. The State of Secrets Sprawl 2025, GitGuardian, 2025
  6. Average global data breach cost now $4.44 million (IBM Cost of a Data Breach Report 2025), Help Net Security, 2025
  7. Logging Cheat Sheet, OWASP Cheat Sheet Series, 2025
  8. OWASP Top 10:2025 Introduction, OWASP Foundation, 2025
Keep reading

Related research

Ship Container
Cloud & Security · 11 Nov 2026

Cost of a Data Breach 2026: What IBM's Report Means for Small Firms

IBM's 2026 study puts the global average breach at $4.99 million. Here is what drives that cost and which findings a small company can act on.

XDQ Editors11 Nov 2026 · 10 min read
Padlock Lock
Cloud & Security · 9 Nov 2026

Deepfake and AI Fraud Statistics 2026, and Defences for Businesses

What the FBI, IBM and Deloitte data says about deepfake and AI-enabled fraud, and the verification steps that stop it reaching your bank account.

XDQ Editors9 Nov 2026 · 10 min read
Cardboard Box
Cloud & Security · 10 Nov 2026

Website Security Checklist for Small Businesses (2026)

A practical, defensive checklist for keeping a small business website patched, encrypted, backed up and monitored, based on 2026 data.

XDQ Editors10 Nov 2026 · 11 min read
Next step

Want your cloud set up and secured properly from day one?

Send the workflow. Our sales experts reply with a scope and a custom quotation within two working days, or a straight answer that it isn't a fit.

Secure by default · find → fix → verify · authorised scope only
Send a brief →enquiry@xdqlabs.com