Engineer working on a laptop connected to a device test rig, surrounded by wiring harnesses and test hardware

Firmware vulnerabilities & fixes

OWASP API Security Top 10 (2023), explained

Last updated:

The OWASP API Security Top 10 is the 2023 edition of OWASP's list of ten categories of API-specific security risk, distinct from the main OWASP Top 10's broader web-application focus. Per its own introduction, its goal is "to educate those involved in API development and maintenance, for example, developers, designers, architects, managers, or organizations." The first edition was published in 2019; the current, second edition was released 3 July 2023. The document is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0), per the project's GitHub repository.

Key takeaways

  • The OWASP API Security Top 10 is the 2023 edition of OWASP's list of ten categories of API-specific security risk, distinct from the main OWASP Top 10's broader web-application focus.
  • Broadened scope: resource consumption shifted from a narrower focus on request rate toward the fuller range of resources (compute, memory, storage, third-party API costs) a single API call can consume.
  • OWASP's Methodology and Data page is explicit that this is not a purely statistical ranking.
  • For a connected product, the object-level and function-level authorization categories carry the most weight, because the cloud API is what turns a stolen or guessed identifier into control of someone else's hardware.

What are the 10 categories in the 2023 edition?

Risk Name OWASP's own description
API1:2023 Broken Object Level Authorization "APIs tend to expose endpoints that handle object identifiers, creating a wide attack surface of Object Level Access Control issues."
API2:2023 Broken Authentication "Authentication mechanisms are often implemented incorrectly, allowing attackers to compromise authentication tokens or exploit implementation flaws."
API3:2023 Broken Object Property Level Authorization "Lack of or improper authorization validation at the object property level leads to information exposure or manipulation by unauthorized parties."
API4:2023 Unrestricted Resource Consumption "Successful attacks can lead to Denial of Service or an increase of operational costs through excessive consumption of APIs."
API5:2023 Broken Function Level Authorization "Complex access control policies with unclear separation between administrative and regular functions tend to lead to authorization flaws."
API6:2023 Unrestricted Access to Sensitive Business Flows "APIs expose business flows without compensating for how functionality could harm the business if used excessively in an automated manner."
API7:2023 Server Side Request Forgery "SSRF flaws occur when an API fetches remote resources without validating user-supplied URIs, enabling crafted malicious requests."
API8:2023 Security Misconfiguration "Complex configurations meant to customize APIs can be missed or mishandled by engineers, opening doors to various attack vectors."
API9:2023 Improper Inventory Management "APIs expose more endpoints than traditional web applications, making proper documentation and inventory of hosts and versions critical."
API10:2023 Unsafe Consumption of APIs "Developers adopt weaker security standards for third-party API data than user input, making integrated services attractive attack targets."

What changed from the 2019 edition?

OWASP's own release notes describe the 2023 edition arriving "exactly four years after" the 2019 first release, with several structural changes:

  • Merged: 2019's Excessive Data Exposure and Mass Assignment were combined into API3:2023 Broken Object Property Level Authorization — the release notes state the merge targets their "common root cause: object property level authorization validation failures."
  • New: API6:2023 Unrestricted Access to Sensitive Business Flows, covering threats like scalping and fake-account creation that a narrow rate-limiting discussion didn't capture, and API10:2023 Unsafe Consumption of APIs, covering risk from trusting data returned by third-party or integrated APIs.
  • Broadened scope: resource consumption shifted from a narrower focus on request rate toward the fuller range of resources (compute, memory, storage, third-party API costs) a single API call can consume.

Three out of the top five 2023 entries involve authorization in some form (object level, object property level, and function level) — a theme OWASP's own announcement attributes to increasingly complex APIs exposing thousands of endpoints, each a potential access-control decision point.

How was the 2023 ranking produced?

OWASP's Methodology and Data page is explicit that this is not a purely statistical ranking. The project ran a public Call for Data from September to November 2022, but states plainly: "this call for data did not result in data that would have enabled a relevant statistical analysis." The team instead combined a review of publicly available API security incident data from 2019–2022 with internal and community discussion of emerging trends, then applied the OWASP Risk Rating Methodology — with "prevalence ratings decided from a consensus among the project team members, based on their experience in the field" — before the draft went through expert and community review as a Release Candidate. This matters for how an organization should use the list: it is a curated, expert-driven snapshot of what practitioners see as the most important API-specific risks, not a reproducible statistical frequency count.

Which categories hit a device cloud API hardest?

For a connected product, the object-level and function-level authorization categories carry the most weight, because the cloud API is what turns a stolen or guessed identifier into control of someone else's hardware. A device back-end usually has three API surfaces: the device-facing API (telemetry, configuration pull, OTA update checks), the companion-app API (pairing, remote control, sharing a device with family members) and an installer or support API (fleet views, remote diagnostics). The table maps each category to the shape it tends to take there, with the first control from OWASP's own "How To Prevent" text.

Category How it shows up in a device back-end First control (per the OWASP category page)
API1 BOLA /devices/{serial}/unlock accepts any serial number, not only devices bound to the caller's account Check, in every function that takes a client-supplied ID, that the logged-in user may perform that action on that record
API2 Broken Authentication A shared API key baked into every firmware image doubles as the device's identity OWASP: "API keys should not be used for user authentication. They should only be used for API clients authentication." Use per-device credentials
API3 Property-level authorization A shared-user PATCH on a device sets owner_id or firmware_channel Allow changes only to the properties the client should update; avoid automatic binding of input to internal objects
API4 Resource consumption Unlimited upload size on crash-dump or log upload endpoints; paid SMS alerts with no cap Set execution timeouts, upload size limits and per-client operation limits
API5 Function-level authorization Installer or factory endpoints (/admin/devices/reassign) reachable with a consumer token Deny by default; check role on every administrative function
API6 Sensitive business flows Scripted device-claim or trial-activation flows Identify the flows that hurt the business if automated, then pick protections for those flows
API7 SSRF A user-configurable webhook or "custom firmware URL" fetched by the cloud Validate and restrict client-supplied URLs before the server fetches them
API8 Misconfiguration Verbose stack traces from the device gateway; missing TLS on a legacy MQTT bridge Harden every layer; send no stack traces; enforce TLS
API9 Inventory /v1/ kept alive for old firmware that never updated, without the newer checks Keep an inventory of hosts and versions and a retirement plan per version
API10 Unsafe consumption The cloud trusts a voice-assistant or energy-tariff partner API without validation or timeouts Validate third-party data like user input; set timeouts; do not follow redirects blindly

OWASP's own API1 page includes a device case: a car maker's mobile API that starts and stops the engine and locks and unlocks doors from a VIN, where "the API fails to validate that the VIN represents a vehicle that belongs to the logged in user." Swap the VIN for a lock serial number or a camera MAC address and the same flaw applies to most consumer and industrial device clouds.

What does object-level authorization look like when the object is a device?

It means the server checks a binding between the authenticated principal and the specific device record on every request, rather than trusting that the caller only knows identifiers it should know. Device identifiers are rarely secret: serial numbers are printed on labels and packaging, MAC addresses are broadcast over Wi-Fi and Bluetooth, and sequential IDs are easy to enumerate. OWASP's API1 page warns that comparing the session's user ID with a parameter "isn't a sufficient solution" and covers "only a small subset of cases"; the check has to follow the user policies and hierarchy.

For a device cloud that hierarchy usually has four levels: account owner, shared users (with their own permission set), installer or support staff (time-bound, scoped to a site), and the device itself. A practical pattern:

  1. Store an explicit device_bindings relation (principal, device, role, expiry) and resolve it server-side for each call.
  2. Authorise the action, not only the device: a shared user may read status but not reassign ownership or change the update channel.
  3. Treat the device credential as a separate principal that can only reach its own record and its own update metadata.
  4. Use random, unguessable record IDs where possible, as OWASP's API1 page recommends, while still enforcing the binding check, because labels leak.
  5. Write tests that call each endpoint with a second account's token and a device serial number from the first account, and block releases when they pass. OWASP's API1 prevention list says "Do not deploy changes that make the tests fail."

The category maps to CWE-639, "Authorization Bypass Through User-Controlled Key," which is a useful tag when the finding lands in a vulnerability tracker or a security advisory.

How should a device without a browser authenticate to the API?

Give each device its own credential and use a standard grant for user linking, rather than a shared API key. OWASP's API2 guidance is blunt: "Don't reinvent the wheel in authentication, token generation, or password storage. Use the standards." For devices that have a network connection but no browser or keyboard, the IETF standard is the OAuth 2.0 Device Authorization Grant, RFC 8628 (August 2019): the device shows a short user_code, the user approves it on a phone or laptop, and the device polls for its token.

That flow has its own API risk. RFC 8628 §5.1 notes that user codes are short for usability and therefore low in entropy, recommends that "the server rate-limit user code attempts," and works through an example: an 8-character base-20 code carries about 34.5 bits, so the rate limit and validity period should allow only around 5 guesses for a 2^-32 chance of success. A successful guess links the victim's device to the attacker's account. Treat the verification endpoint like a login endpoint, which is also what OWASP's API2 page asks for credential-recovery flows. For the wider OAuth deployment, the IETF's Best Current Practice for OAuth 2.0 Security, RFC 9700 (BCP 240, January 2025) is the current reference. The companion app's side of the login is covered in our OWASP MASVS guide, and assurance levels for user logins in our NIST SP 800-63 guide.

Worked example: how would you review a smart-lock cloud API?

Synthetic example. The product, endpoints and findings below are invented for illustration and do not describe any real vendor.

A team ships a Wi-Fi smart lock with a companion app and a cloud API at api.example-lock.test. Before a firmware release that adds guest access codes, they walk the API against the 2023 list:

  1. Inventory (API9)They export every route from the gateway config and find /v1/ still answers. It exists for locks on firmware 1.x, about one in ten units in the field by their telemetry. /v1/locks/{serial}/state has no binding check. They add the check to /v1/ rather than waiting for the fleet to update, and set a retirement date.
  2. Object level (API1)With two test accounts, account B calls POST /v2/locks/{serial-of-A}/unlock. It returns 403. They repeat it for every route that takes a serial, including the new guest-codes route, which returns 200. They fix it before release.
  3. Property level (API3)PATCH /v2/locks/{serial} accepts {"owner_account": "B"} from a shared user. They move to an allow-list of writable fields per role.
  4. Function level (API5)The support tool's /internal/reassign is reachable from the public gateway with any valid token. They move it behind the staff identity provider.
  5. Authentication (API2)The device-linking code is 6 digits with no attempt limit. Following RFC 8628 §5.1, they add a per-code attempt cap and a 10-minute expiry.
  6. Consumption (API4 / API10)The doorbell-to-SMS integration has no daily cap, and the partner's response is stored unvalidated. They add a per-account cap, a timeout and schema validation.

The release ships with tests for steps 2 to 5 in CI, so a regression fails the build.

What are the common pitfalls?

  • Treating the device network as trusted. A request with a valid device certificate can still ask for another device's record. Bind the certificate to one device ID on the server side.
  • Fixing only the newest API version. Old firmware keeps old endpoints alive for years. API9 is where most device fleets carry their oldest flaws.
  • Using the serial number as a secret. It is printed on the box. Treat it as an identifier only.
  • Rate-limiting the login but not the linking or recovery flow. OWASP's API2 page asks for brute-force protection on recovery endpoints, and RFC 8628 asks for it on user codes.
  • Leaving the API out of the product security scope. The cloud API is part of the product. The OWASP IoT Top 10 lists "Insecure Ecosystem Interfaces" for this reason, and the OWASP ASVS gives testable requirements for the back-end code.
  • Missing the back-end's own components. API8's first OWASP scenario is a logging library with JNDI lookups enabled by default. That is the pattern behind CVE-2021-44228, and it shows that the API server's dependencies need the same vulnerability tracking as the firmware.

Where does SoftWSP fit?

SoftWSP does not scan or penetration-test APIs. Its part is the component side of the same product. It extracts a software bill of materials from a firmware image as CycloneDX 1.6 JSON. It matches the components against NVD, the CISA KEV catalogue, FIRST EPSS scores and CISA ICS advisories. It then drafts a cited report for an engineer to review. That covers the device half of the API8 "out of date systems" problem, so the authorization review above can concentrate on the API logic. See firmware SBOM, the free SBOM viewer, and device firmware security for scoped engineering work.

This guide summarises a public OWASP document; it does not constitute legal or compliance advice.

Put this guide to work

Frequently asked questions

What is the OWASP API Security Top 10?

An awareness document published by the OWASP API Security Project. Per its own introduction, "the primary goal of the OWASP API Security Top 10 is to educate those involved in API development and maintenance, for example, developers, designers, architects, managers, or organizations" about the most important security risks specific to APIs. The first edition was published in 2019; the current edition, released 3 July 2023, is the second. It is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0), per the project's GitHub repository.

What are the 10 categories in the 2023 edition?

Per OWASP's own Top 10 API Security Risks – 2023 page: API1:2023 Broken Object Level Authorization, API2:2023 Broken Authentication, API3:2023 Broken Object Property Level Authorization, API4:2023 Unrestricted Resource Consumption, API5:2023 Broken Function Level Authorization, API6:2023 Unrestricted Access to Sensitive Business Flows, API7:2023 Server Side Request Forgery, API8:2023 Security Misconfiguration, API9:2023 Improper Inventory Management, and API10:2023 Unsafe Consumption of APIs.

What changed between the 2019 and 2023 editions?

OWASP's own release notes describe the 2023 edition as arriving "exactly four years after" the 2019 edition's first release. The two 2019 categories Excessive Data Exposure and Mass Assignment were merged into the new API3:2023 Broken Object Property Level Authorization, which the release notes describe as focusing on their "common root cause: object property level authorization validation failures." Two categories are entirely new: API6:2023 Unrestricted Access to Sensitive Business Flows (addressing threats like scalping and fake-account creation that rate-limiting discussions alone didn't capture) and API10:2023 Unsafe Consumption of APIs (risk from trusting data returned by third-party/integrated APIs). The remaining categories carried forward with updated detail, and resource consumption shifted from a narrow focus on request rate to the broader scope of what resources an API call can consume.

How was the 2023 ranking produced — is it data-driven like the main OWASP Top 10?

OWASP's own Methodology and Data page states the project ran a public Call for Data from September to November 2022, but reports plainly that "this call for data did not result in data that would have enabled a relevant statistical analysis." The 2023 list is therefore primarily expert-consensus based: the project team reviewed publicly available API security incident data from 2019–2022, discussed emerging trends internally and with the community, and applied the OWASP Risk Rating Methodology, with "prevalence ratings decided from a consensus among the project team members, based on their experience in the field," before the draft went through expert and community review as a Release Candidate.

Sources

  1. OWASP API Security Top 10 2023 — Table of Contents
  2. OWASP API Security Top 10 2023 — Introduction
  3. OWASP API Security Top 10 2023 — OWASP Top 10 API Security Risks – 2023
  4. OWASP API Security Top 10 2023 — Release Notes
  5. OWASP API Security Top 10 2023 — Methodology and Data
  6. OWASP blog: OWASP API Security Top 10 2023 has been released (3 July 2023)
  7. GitHub: OWASP/API-Security (license, project leaders)
  8. OWASP API Security Top 10 2023 — API1:2023 Broken Object Level Authorization
  9. OWASP API Security Top 10 2023 — API2:2023 Broken Authentication
  10. OWASP API Security Top 10 2023 — API3:2023 Broken Object Property Level Authorization
  11. OWASP API Security Top 10 2023 — API4:2023 Unrestricted Resource Consumption
  12. OWASP API Security Top 10 2023 — API6:2023 Unrestricted Access to Sensitive Business Flows
  13. OWASP API Security Top 10 2023 — API7:2023 Server Side Request Forgery
  14. OWASP API Security Top 10 2023 — API8:2023 Security Misconfiguration
  15. OWASP API Security Top 10 2023 — API9:2023 Improper Inventory Management
  16. OWASP API Security Top 10 2023 — API10:2023 Unsafe Consumption of APIs
  17. IETF RFC 8628: OAuth 2.0 Device Authorization Grant (August 2019)
  18. IETF RFC 9700: Best Current Practice for OAuth 2.0 Security (BCP 240, January 2025)
  19. MITRE CWE-639: Authorization Bypass Through User-Controlled Key
  20. NVD: CVE-2021-44228
Was this guide helpful?

In this cluster