How To Control Access To REST APIs

hackerExposing your data or application through a REST API is a wonderful way to reach a wide audience.

The downside of a wide audience, however, is that it’s not just the good guys who come looking.

Securing REST APIs

Security consists of three factors:

  1. Confidentiality
  2. Integrity
  3. Availability

In terms of Microsoft’s STRIDE approach, the security compromises we want to avoid with each of these are Information Disclosure, Tampering, and Denial of Service. The remainder of this post will only focus on Confidentiality and Integrity.

In the context of an HTTP-based API, Information Disclosure is applicable for GET methods and any other methods that return information. Tampering is applicable for PUT, POST, and DELETE.

Threat Modeling REST APIs

A good way to think about security is by looking at all the data flows. That’s why threat modeling usually starts with a Data Flow Diagram (DFD). In the context of a REST API, a close approximation to the DFD is the state diagram. For proper access control, we need to secure all the transitions.

The traditional way to do that, is to specify restrictions at the level of URI and HTTP method. For instance, this is the approach that Spring Security takes. The problem with this approach, however, is that both the method and the URI are implementation choices.

link-relationURIs shouldn’t be known to anybody but the API designer/developer; the client will discover them through link relations.

Even the HTTP methods can be hidden until runtime with mature media types like Mason or Siren. This is great for decoupling the client and server, but now we have to specify our security constraints in terms of implementation details! This means only the developers can specify the access control policy.

That, of course, flies in the face of best security practices, where the access control policy is externalized from the code (so it can be reused across applications) and specified by a security officer rather than a developer. So how do we satisfy both requirements?

Authorizing REST APIs

I think the answer lies in the state diagram underlying the REST API. Remember, we want to authorize all transitions. Yes, a transition in an HTTP-based API is implemented using an HTTP method on a URI. But in REST, we shield the URI using a link relation. The link relation is very closely related to the type of action you want to perform.

The same link relation can be used from different states, so the link relation can’t be the whole answer. We also need the state, which is based on the representation returned by the REST server. This representation usually contains a set of properties and a set of links. We’ve got the links covered with the link relations, but we also need the properties.

PolicyIn XACML terms, the link relation indicates the action to be performed, while the properties correspond to resource attributes.

Add to that the subject attributes obtained through the authentication process, and you have all the ingredients for making an XACML request!

There are two places where such access control checks comes into play. The first is obviously when receiving a request.

You should also check permissions on any links you want to put in the response. The links that the requester is not allowed to follow, should be omitted from the response, so that the client can faithfully present the next choices to the user.

Using XACML For Authorizing REST APIs

I think the above shows that REST and XACML are a natural fit.

All the more reason to check out XACML if you haven’t already, especially XACML’s REST Profile and the forthcoming JSON Profile.

Book Review: The Security Development Lifecycle (SDL)

In The Security Development Lifecycle (SDL), A Process for Developing Demonstrably More Secure Software, authors Michael Howard and Steven Lipner explain how to build secure software through a repeatable process.

The methodology they describe was developed at Microsoft and has led to a measurable decrease in vulnerabilities. That’s why it’s now also used elsewhere, like at EMC (my employer).

Chapter 1, Enough is Enough: The Threats have Changed, explains how the SDL was born out of the Trustworthy Computing initiative that started with Bill Gates’ famous email in early 2002. Most operating systems have since become relatively secure, so hackers have shifted their focus to applications and the burden is now on us developers to crank up our security game. Many security issues are also privacy problems, so if we don’t, we are bound to pay the price.

Chapter 2, Current Software Development Methods Fail to Produce Secure Software, reviews current software development methods with regard to how (in)secure the resulting applications are. It shows that the adage given enough eyeballs, all bugs are shallow is wrong when it comes to security. The conclusion is that we need to explicitly include security into our development efforts.

Chapter 3, A Short History of the SDL at Microsoft, describes how security improvement efforts at Microsoft evolved into a consistent process that is now called the SDL.

Chapter 4, SDL for Management, explains that the SDL requires time, money, and commitment from senior management to prioritize over time to market. We’re talking real commitment, like delaying the release of an insecure application.

Chapter 5, Stage 0: Education and Awareness, starts the second part of the book, that describes the stages of the SDL. It all starts with educating developers about security. Without this, there’s no real chance of delivering secure software.

Chapter 6, Stage 1: Project Inception, sets the security context for the development effort. This includes assigning someone to guide the team through the SDL, building security leaders within the team, and setting up security expectations and tools.

Chapter 7, Stage 2: Define and Follow Best Practices, lists common secure design principles and describes attack surface analysis and attack surface reduction. The latter is about reducing the amount of code accessible to untrusted users, for example by disabling certain features by default.

Chapter 8, Stage 3: Product Risk Assessment, shows how to determine the application’s level of vulnerability to attack and its privacy impact. This helps to determine what level of security investment is appropriate for what parts of the application.

Chapter 9, Stage 4: Risk Analysis, explains threat modeling. The authors think that this is the practice with the most significant contribution to an application’s security. The idea is to understand the potential threats to the application, the risks those threats pose, and the mitigations that can reduce those risks. Threat models also help with code reviews and penetration tests. The chapter uses a pet shop website as an example.

[Note that there is now a tool that helps you with threat modeling. In this tool, you draw data flow diagrams, after which the tool uses the STRIDE approach to automatically find threats. The tool requires Visio 2007+.]

Chapter 10, Stage 5: Creating Security Documents, Tools, and Best Practices for Customers, describes the collateral that helps customers install, maintain, and use your application securely.

Chapter 11, Stage 6: Secure Coding Policies, explains the need for prescribing security-specific coding practices, educating developers about them, and verifying that they are adhered to. This is a high-level chapter, with details following in later chapters.

Chapter 12, Stage 7: Secure Testing Policies, describes the various forms of security testing, like fuzz testing, penetration testing, and run-time verification.

Chapter 13, Stage 8: The Security Push, explains that the goal of a security push is to hunt for security bugs and triage them. Fixes should follow the push. A security push doesn’t really fit into the SDL, since the goal is to prevent vulnerabilities. It can, however, be useful for legacy (i.e. pre-SDL) code.

Chapter 14, Stage 9: The Final Security Review, describes how to assess (from a security perspective) whether the application is ready to ship. A questionnaire is filled out to show compliance with the SDL, the threat models are reviewed, and unfixed security bugs are reviewed to make sure none are critical.

Chapter 15, Stage 10: Security Response Planning, explains that you need to be prepared to respond to the discovery of vulnerabilities in your deployed application, so that you can prevent panic and follow a solid process. You should have a Security Response Center outside your development team that interfaces with security researchers and others who discover vulnerabilities and guides the development team through the process of releasing a fix. It’s also important to feed back lessons learned into the development process.

Chapter 16, Stage 11: Product Release, explains that the actual release is a non-event, since all the hard work was done in the Final Security Review.

Chapter 17, Stage 12: Security Response Execution, describes the real-world challenges associated with responding to reported vulnerabilities, including when and how to deviate from the plan outlined in Security Response Planning. Above all, you must take the time to fix the root problem properly and to make sure you’re not introducing new bugs.

Chapter 18, Integrating SDL with Agile Methods, starts the final part of the book. It shows how to incorporate agile practices into the SDL, or the other way around.

Chapter 19, SDL Banned Function Calls, explains that some functions are so bad from a security perspective, that they never should be used. This chapter is heavily focused on C.

Chapter 20, SDL Minimum Cryptographic Standards, gives guidance on the use of cryptography, like never roll your own, make the use of crypto algorithms configurable, and what key sizes to use for what algorithms.

Chapter 21, SDL-Required Tools and Compiler Options, describes security tools you should use during development. This chapter is heavily focused on Microsoft technologies.

Chapter 22, Threat Tree Patterns, shows a number of threat trees that reflect common attack patterns. It follows the STRIDE approach again.

The appendix has information about the authors.

I think this book is a must-read for every developer who is serious about building secure software.

Abuse Cases

Gary McGraw describes several best practices for building secure software. One is the use of so-called abuse cases. Since his chapter on abuse cases left me hungry for more information, this post examines additional literature on the subject and how to fit abuse cases into a Security Development Lifecycle (SDL).

Modeling Functional Requirements With Use Cases

Abuse cases are an adaptation of use cases, abstract episodes of interaction between a system and its environment.

A use case consists of a number of related scenarios. A scenario is a description of a specific interaction between the system and particular actors. Each use case has a main success scenario and some additional scenarios to cover variations and exceptional cases.

Actors are external agents, and can be either human or non-human.

For better understanding, each use case should state the goal that the primary actor is working towards.

Use cases are represented in UML diagrams (see example on left) as ovals that are connected to stick figures, which represent the actors. Use case diagrams are accompanied by textual use case descriptions that explain how the actors and the system interact.

Modeling Security Requirements With Abuse Cases

An abuse case is a use case where the results of the interaction are harmful to the system, one of the actors, or one of the stakeholders in the system. An interaction is harmful if it decreases the security (confidentiality, integrity, or availability) of the system.

Abuse cases are also referred to as misuse cases, although some people maintain they’re different. I think the two concepts are too similar to treat differently, so whenever I write “abuse case”, it refers to “misuse case” as well.

Some actors in regular use cases may also act as attacker in an abuse case (e.g. in the case of an insider threat). We should then introduce a new actor to avoid confusion (potentially using inheritance). This is consistent with the best practice of having actors represent roles rather than actual users.

Attackers are described in more detail than regular actors, to make it easier to look at the system from their point of view. Their description should include the resources at their disposal, their skills, and their objectives.

Note that objectives are longer term than the (ab)use case’s goal. For instance, the attacker’s goal for an abuse case may be to gain root privileges on a certain server, while her objective may be industrial espionage.

Abuse cases are very different from use cases in one respect: while we know how the actor in a use case achieves her goal, we don’t know precisely how an attacker will break the system’s security. If we would, we would fix the vulnerability! Therefore, abuse case scenarios describe interactions less precisely than regular use case scenarios.

Modeling Other Non-Functional Requirements

Note that since actors in use cases needn’t be human, we can employ a similar approach to abuse cases with actors like “network failure” etc. to model non-functional requirements beyond security, like reliability, portability, maintainability, etc.

For this to work, one must be able to express the non-functional requirement as an interactive scenario. I won’t go into this topic any further in this post.

Creating Abuse Cases

Abuse case models are best created when use cases are: during requirements gathering. It’s easiest to define the abuse cases after the regular use cases are identified (or even defined).

Abuse case modeling requires one to wear a black hat. Therefore, it makes sense to invite people with black hat capabilities, like testers and network operators or administrators to the table.

The first step in developing abuse cases is to find the actors. As stated before, every actor in a regular use case can potentially be turned into an malicious actor in an abuse case.

We should next add actors for different kinds of intruders. These are distinguished based on their resources and skills.

When we have the actors, we can identify the abuse cases by determining how they might interact with the system. We might identify such malicious interactions by combining the regular use cases with attack patterns.

We can find more abuse cases by combining them systematically and recursively with regular use cases.

Combining Use Cases and Abuse Cases

Some people keep use cases and abuse cases separate to avoid confusion. Others combine them, but display abuse cases as inverted use cases (i.e. black ovals with white text, and actors with black heads).

The latter approach makes it possible to relate abuse cases to use cases using UML associations. For instance, an abuse case may threaten a use case, while a use case might mitigate an abuse case. The latter use case is also referred to as a security use case. Security use cases usually deal with security features.

Security use cases can be threatened by new abuse cases, for which we can find new security use cases to mitigate, etc. etc. In this way, a “game” of play and counterplay enfolds that fits well in a defense in depth strategy.

We should not expect to win this “game”. Instead, we should make a good trade-off between security requirements and other aspects of the system, like usability and development cost. Ideally, these trade-offs are made clearly visible to stakeholders by using a good risk management framework.

Reusing Abuse Cases

Use cases can be abstracted into essential use cases to make them more reusable. There is no reason we couldn’t do the same with abuse cases and security use cases.

It seems to me that this not just possible, but already done. Microsoft’s STRIDE model contains generalized threats, and its SDL Threat Modeling tool automatically identifies which of those are applicable to your situation.

Conclusion

Although abuse cases are a bit different from regular use cases, their main value is that they present information about security risks in a format that may already be familiar to the stakeholders of the software development process. Developers in particular are likely to know them.

This should make it easier for people with little or no security background to start thinking about securing their systems and how to trade-off security and functionality.

However, it seems that threat modeling gives the same advantages as abuse cases. Since threat modeling is supported by tools, it’s little wonder that people prefer that over abuse cases for inclusion in their Security Development Lifecycle.