In line with MITRE, the commonest type of API (software programming interface) misuse happens when the caller doesn’t honor its finish of a contract. Within the context of this text, a “contract” refers to a proper, exact settlement that outlines the anticipated behaviors, inputs, outputs, and unintended effects that an API ensures to any caller, making certain that each the API and its shoppers adhere to specified constraints and usages. This idea is essential in stopping misuse by clearly defining the boundaries and necessities for each events concerned within the interplay. This weblog publish explores contract programming and particularly how that applies to the constructing, upkeep, and safety of APIs.
API misuse usually happens as a result of not understanding the state of the system behind an API, which can result in incorrect ordering of calls and finish in an error state that could be a vulnerability. This misuse can even occur when an implementation of an API doesn’t meet the specification. For instance, a shopper could also be anticipating a sure output per the specification however obtain one thing totally different. Lastly, misuse of an API can occur in an object-oriented programming (OOP) context with particular subclass implementations. These implementations could not present the identical performance that’s mandated by the tremendous class or interface. Within the design and implementation of software program methods there exists a principle of contracts that may assist to resolve a few of these points.
An API is, in a normal sense, a contract between a supplier of a software program and the patron of that software program about what the system will do. This concept of contract programming or Design by Contract was coined by Bertrand Meyer in 1986. On this paradigm, a software program engineer defines formally the specification for every perform or methodology that the system exposes ((within the context of this paper, the phrases “perform” and “methodology” are used interchangeably). This specification contains noting pre-conditions, post-conditions, and invariants. Whereas usually design apply for enhancing the verifiability of methods, this contract programming assemble additionally allows API safety.
Pre- and Postconditions of a Contract
We outline a features contract because the set of pre- and post- circumstances and the invariants of the perform that should maintain.
Preconditions are the set of standards required earlier than a perform may be executed. These are issues that the service or API supplier count on to be true earlier than a perform is known as. An instance of this, within the context of an API, is {that a} precondition for accessing a protected endpoint be that the caller supplies a legitimate authentication token. One other instance is a perform that requires a legitimate (i.e., not null) pointer be handed to it. In both of those circumstances, if the precondition just isn’t met (i.e., the token is invalid, or the pointer is null), then the contract is damaged.
Postconditions are the state or set of standards that should be true after a perform is executed. Postconditions for an API would be the return of some specified information and an HTTP 200 standing code. A caller or shopper that makes use of an API perform whose preconditions aren’t met just isn’t entitled to the postconditions. The system that’s furnishing the API is predicted to offer the publish circumstances. Lastly, invariants are the info or state that can’t be modified by perform execution whatever the operation or transformation utilized by the perform.
Subsequently, to honor the contract means to reply the three questions of a Hoare triple:
- What does the contract count on?
- What does the contract assure?
- What does the contract preserve?
Defining an API Interface
As an interface, an API sometimes is outlined in information definition language (DDL), interface description language (IDL), or simply plain textual content. Consequently, an interface’s implementation will not be true to the specification. Formal strategies present a way of verifying that an implementation refines a specification. Making certain an implementation meets all expectations of an interface can be carefully tied to the Liskov Substitution Precept. In discussing each refinement of a specification and the Liskov substitution precept we are able to generalize the next constraints for a perform:
- Preconditions can’t be strengthened (i.e., an implementation could not settle for a narrower vary of enter than the specification dictates). For instance, an implementation of a pop() methodology on a Stack can not add a precondition that the stack will need to have a minimal dimension of 5 parts earlier than permitting a name to pop().
- Postconditions can’t be weakened (i.e., an implementation could not return a bigger vary of output than the specification dictates). For instance, after calling push(ingredient) on a Stack the stack should mirror the addition of precisely one new ingredient, however no more.
- Invariants can’t be weakened (i.e., an implementation could not alter the state of invariants listed within the specification). For instance, the dimensions of a stack must not ever be damaging, whatever the variety of pop() or push() operations carried out.
Along with errors on the patron facet of an API, errors can be brought on by not absolutely implementing the interface of the API or doing so incorrectly. For instance, the Open Worldwide Utility Safety Challenge (OWASP) basis publishes a listing of the prime 10 API safety dangers. For 2023 the highest threat was Damaged Object Degree Authorization (BOLA). BOLA is an instance of an implementation not honoring a contract precondition, resembling a request to a given API perform or endpoint should comprise an authorization token that’s legitimate for the actual object being requested.
Who Ought to Verify the Pre-and Postconditions?
This query relies on the type and structure of the codebase that’s implementing an API. In lots of circumstances the supplier of the API would require robust preconditions and won’t even try and work if they aren’t met. This constraint places the burden on the consumer to make sure that all the things is legitimate and within the correct state earlier than calling an API. However, the methods of defensive programming recommend that it’s probably higher to deal with unexpected circumstances extra gracefully. Meyer suggests when designing by contract that dealing with one case nicely and requiring robust preconditions is a greatest apply that has proved profitable.
Programming Language Instruments for Defining API Contracts
How contracts are outlined in a specific language varies. In Java using Javadoc feedback to doc the parameters, return worth, exceptions, and the features goal is a typical (although much less formal) method of documenting a contract. There are additionally quite a lot of instruments that provide various ranges of ritual for outlining contracts that may assist to allow verification of API utilization. Some notable examples are:
- Eiffel
- Java by way of Java Modeling Language (JML)
- Kotlin (natively)
- Rust by way of the contracts crate
- Ada 2012 (natively)
- API Blueprint
- OpenAPI
Given the prevalence of HTTP-based REST APIs, OpenAPI is a related instrument and format for specifying endpoints, enter and output for every operation, authentication strategies, and different data. The usage of the OpenAPI specification to outline an API aligns nicely with the design-by-contract paradigm of specifying the preconditions, such because the area of inputs and an outline of the endpoint. OpenAPI additionally permits for specifying the return from an endpoint together with the return code, an outline of what that return code means, the schema of any returned information, and examples of the info.
Particularly, within the realm of API safety, OpenAPI additionally permits for specification of the authentication and authorization necessities for every endpoint. In the documentation, OpenAPI refers to this as a safety scheme. Within the 3.0 model, this safety scheme contains HTTP authentication, API Keys, OAuth2, and OpenID Join.
Doubtlessly a spot that OpenAPI falls brief is within the skill to specify invariants of a perform. As an illustration, in a REST API, GET requests ought to be idempotent. There’s, nonetheless, no strategy to doc exterior of a textual content description what an endpoint could or could not change by way of state.
Whereas Open API and the opposite listed instruments all supply a machine readable or parseable format, as beforehand talked about, even a textual content description of a features contract might help. The benefit of a machine-readable contract, nonetheless, is the flexibility to generate take a look at circumstances for the contract.
There are a number of open-source instruments, resembling RESTler and Dredd, that can eat an OpenAPI spec and robotically generate and execute take a look at circumstances towards an implementation. Equally with Java and the Java Modeling Language (JML), there are functions that may remodel the Javadoc feedback into runtime assertions. An instance of this method is the JML compiler that provides in assertions to the Java bytecode.
Advantages to API Testing
As now we have explored, there are various instruments for supporting contract programming. Nevertheless, these instruments include a price. Specifically, builders should be educated on their use; the instruments should be built-in right into a product’s DevSecOps pipeline, and they’re yet one more dependency that should be maintained and up to date. Along with the patron advantages of offering a contract, what advantages can the API builders get from utilizing these instruments? I contend that the most important benefit to builders working below the contract programming paradigm is the flexibility to check the interface with out testing the implementation.
Josh Bloch, CMU professor and previously of Google and Solar Microsystems wrote, “Code the use-cases towards your API earlier than you implement it.” A product with a well-defined contract allows the group to check out an API specification and write instance consumer code that makes use of the features or endpoints very early on within the growth cycle. This method eliminates any time spent implementing a selected perform after which discovering out the perform is awkward or laborious to make use of kind the consumer perspective.
This idea additionally extends to integration testing of various software program modules. For big, complicated methods it may be laborious to assemble all of the customers to carry out dwell testing of every part. Equally, some methods can show laborious to simulate a take a look at surroundings for. Maybe the goal system is extremely costly to function on (resembling quantum computer systems at ~ $1.60 per second) or the system just isn’t even constructed but. In each circumstances having a contract that precisely represents a software program module or library can assist the mixing testing carried out by each producers and shoppers of the software program.
Growing API Usability Will increase Safety
Whereas APIs can be utilized by each people and different functions, they’re in the end designed and carried out by people. Ignoring the usability or developer expertise of an API can result in safety issues whereas growing API usability can bolster safety. For instance, a research by Sascha Fahl et. al discovered that in 13,500 standard free Android apps, eight p.c had misused the APIs for the Safe Sockets Layer (SSL) or its successor, the Transport Layer Safety (TLS), and have been thus susceptible to man-in-the-middle and different assaults. A follow-on research of Apple iOS apps discovered that 9.7 p.c have been susceptible with causes together with vital difficulties utilizing safety APIs accurately. The authors of the research advocate quite a few modifications that may improve the usability and safety of the APIs.
Brad Myers contends that API safety is a perform of unhealthy code written by programmers who’re human. Simpler-to-use APIs due to this fact assist safety by making good code simpler to write down and unhealthy code more durable. To help this method, contract pushed programming could be a means to ease the burden of counting on documentation exterior of the supply code as a result of it has been proven that many software program builders desire to make use of supply code over official documentation.
Each API doesn’t present supply code. Nevertheless, even for these which might be absolutely open, centralizing the API guidelines and expectations inside a contract might help streamline the developer expertise. This idea of a code-driven method to studying meshes nicely with the truth that most contract programming mechanisms are immediately embedded withing the supply code that implements the contract. Having a transparent, easy-to-find and easy-to-use API contract can stop unintentional misuse.
One other instance of a damaged contract that had safety implications is Heartbleed. Within the implementation of OpenSSL, the heartbeat request message might be exploited to overread the buffer when asking for extra information than the payload wanted. This exploit was a violation of the contract within the sense that the payload_length discipline ought to have been the identical because the payload however was not. On reflection, this error is a basic buffer over-read, however it affected many methods. If a contract has explicitly outlined the precondition that the payload and payload size should be the identical, the error could have been extra apparent to the implementer. Whereas there are different means to resolve this identical drawback by means of automated code restore or utilizing languages with extra strict compilers, contract pushed programming may present a language agnostic strategy to keep away from related errors in implementation.
The Way forward for API Contract Programming
Contract programming within the context of APIs is a strong idea that may assist guarantee an API conforms to a specification. APIs by their nature symbolize a black field the place an implementation and the how of the system is opaque to the consumer. Given the character of APIs, it is very important inform API customers what precisely is required and what to anticipate. A standardized method to representing these contracts helps testers automate and validate APIs. Nicely-defined contracts can even assist within the developer expertise of an API and supply extra formal verification of methods that require much more assurance.