Best Practices for REST API Testing

What to Expect on This Page

The Challenges of REST API Testing

Delivering high quality REST APIs is a challenging task, as they are highly complex: Some code is difficult to reach, endpoints can consume countless parameter combinations, and constantly communicate with a vast number of other systems. Looking for any type of weakness putting stability at risk or even opening doors for attackers in REST APIs manually is like looking for a needle in a haystack. To deal with the complexity of REST APIs, dev teams test APIs with automated methods. Below, you can find an overview of the top 6 challenges of REST API testing.

1. Initial Documentation of a REST API

As presented below, REST APIs consist of various different parameters such as request method, request URI and query parameter - just to name a few. These parameters can take up countless combinations that have to be tested, as specific parameter combinations can lead to erroneous program states. Documentation like e.g. “OpenAPI” may not even reveal the existence of some parameters.

2. Validating REST API Parameters

Validating REST API parameters is highly challenging. If they are not validated properly, issues such as wrong string/data types and parameter data outside the predefined value range can come up.

3. Evolving the API and Maintaining the Documentation

The documentation is the written contract of the API and specifies how REST APIs handle requests and responses. The challenge in maintaining the documentation is that whenever new parameters are added, parameter limits change or responses change, this has to be reflected in the documentation. Of course, not breaking an existing contract without properly handling it by delivering a breaking change is crucial.

4. Testing REST API Call Sequences (Stateful testing)

Testers must frequently ensure that REST API calls are called in the correct order to test in depth. Even incorrect handling of the order and concurrency issues need to be taken care of as they are frequently found in REST APIs.

5. REST API Testing Set-Up

Setting up automated testing cycles is the part of REST API testing that requires the most manual effort. Especially for large projects, enterprise testing platforms not owned by the developers writing the code are costly to set up and maintain. Testing as close as possible to the time the code is written, will help you to reduce late findings and costly set ups.

6. Error Reporting for REST APIs

Testing the API just from outside without knowledge of the code, error reporting for REST APIs is tricky, as we frequently see a state involved and reproducing might be tricky. The best way to monitor a REST API and report issues is with coverage-guided testing approaches. These approaches ensure knowing each line of code, track where exactly issues happen and even collect coverage information in one go and make reproducing any issue a matter of minutes.

How to Automate Testing for REST APIs

Due to the complex structure of REST APIs, automated testing is one of the most effective ways to ensure smooth, reliable and secure operation. But not all automated testing approaches are equally effective. The fastest way to implement software test automation would be any testing of the final API running on a server with API testing tools, such as Burp or OWASP ZAP, potentially enhanced with some additional system tests. In this setup, typically other, not the teams writing the code take over responsibility for the quality assurance of the API. 

Although these approaches are somewhat automated, they leave plenty of room for improvement, as they still require testers to have prior knowledge about the system under test in order to be effective and to maintain separate tests. These tests are great to test APIs from an attacker's perspective. But they do not make use of the biggest advantage developers of an API have: Knowledge of the source code. They generate test inputs randomly, from static corpora, from OpenAPI imports, or based on heuristics. However, such inputs often fail to reach API specific complex issues and edge cases, since they do not take code coverage into account. For example, such a testing tool would take the API request from above and try out countless different parameter settings in hopes of identifying a request that breaks something.

Automated testing from inside the code is far more effective at finding buggy REST API requests: Since they use information about the source code, these approaches can automatically exclude irrelevant parameter settings from the corpus. Through information about code coverage, they can find crashing REST API requests much faster and much more accurately. This automation also enables better reporting by providing code-coverage visibility. The advantages of this approach are especially useful to test vast microservice environments that are connected through APIs, and projects that are expanding in size.

Testing Before Deploying vs After Deploying

While there are several different ways to test a REST API, one of the most helpful distinctions is to consider testing before and after deploying a service. In this case, this could mean deploying a REST API to a development environment or developer’s workstation, deploying to an internal-only test or staging environment for testing by multiple individuals and teams, or deploying to production to some or all end users. Testing before deployment starts with unit testing, and testing the “happy case”, the underlying business logic of a REST API. Testing against all potential negative cases is highly important at a single API level to keep the number of potential combinations still limited. “Happy case” testing and negative testing can be done in full isolation and with spinning up next related services in a local setup.

Testing after deployment faces some risks and comes with significantly increased cost. It is required but should be replaced with before deployment tests wherever possible. When done right, both testing before and after deployment is critical to ensure a high-quality API.

Testing before deployment is effective for testing business logic and contracts of the service. Functional unit testing is an excellent approach here, since the developer has full control over the source code and can create unit test cases based on the business requirements. Finding any type of bug based on unit testing and integrated end point testing before deployment greatly reduces the risk of such defects affecting end users or even other developer teams that may use the API service. 

Adding edge case or negative testing allows to cover for all unexpected inputs, all that could potentially go wrong on smallest service levels. This ensures robust and reliable systems building on them.

Testing pre-deployment has the benefit of significantly faster execution time, since network connections are mocked out completely. Network overhead is typically the bottle neck when testing APIs. 

Testing the API endpoints before deployment also has the benefit of much faster turnaround times and less management overhead. Fuzz testing is one approach invented to find major software flaws in both the business logic layer and REST API service layer. Using fuzzing to automatically cover all edge cases a developer did not think about and is an excellent way for application developers to “shift left”, take over full responsibility for the API while using familiar test frameworks, tools and processes. It will reduce the number of manually written tests, reduce the time thinking of test cases and even maintain tests continuously.

Fuzz Testing for REST APIs

The most efficient way to implement automated API testing is with feedback-based fuzzing. Feedback-based fuzz testing is a form of data-driven testing, during which code instrumentation is used to measure the test progress within individual microservices and APIs. Through this instrumentation, the fuzzer collects information about what paths the test inputs took in the code, which it then uses to create further inputs that traverse even more code paths. Modern fuzzing solutions can be customized to scan for specific bug classes and to remove blockers. Since this technology can be integrated into any build system, it enables developers to continuously test APIs for all types of bugs caused by unexpected inputs. 

As mentioned above, fuzzing a service pre-deployment and locally on a developer machine like any unit test, is a great approach for testing of a REST API because of the reduced latency of local testing. This allows feedback-based fuzz testing to execute 1000x faster compared to external approaches over the network faster. And it allows tests to run highly efficiently as it tests what is relevant for the code. With the additional benefit of generating code coverage. Teams know exactly what code has been tested and where there are still gaps.

The result: feedback-based fuzzing finds all the code issues that later stages might have found with much higher effort or customers may have found and even worse: hackers might have exploited. Developers can take advantage of fuzzing during development of an API to find all types of defects even before deploying a service to an internal test environment. 

FAQs