> FUNDAMENTALS-5: Minimise the number of permission checks Java is primarily an object-capability language. SecurityManager checks should be considered a last resort.
(Note: quite a lot of Java’s standard library is not designed along object-capability lines so you should take this advice with a pinch of salt).
[1]: https://www.oracle.com/java/technologies/javase/seccodeguide...
It's probably worth posting a quick refresher. The system is old but people don't use it much these days, and the documentation isn't that good. At one point I wrote a small JavaFX PDF viewer that sandboxed the PDF rendering code, to learn the system. I lost the source code apparently, but the hard part wasn't coding it (only a small bit of code was required), it was learning how to configure and use it. I tested the sandbox by opening a PDF that contained an exploit for an old, patched security bug and by using an old, vulnerable version of the PDFbox library. The sandbox successfully stopped the exploit.
Fortunately the Java team still maintain the sandbox and via new technology like the module system and GraalVM, are reinforcing it. In fact, GraalVM introduces a new sandboxing technology as well that's simpler to use than the SecurityManager, however, it's also probably less appropriate for the case of blocking supply chain attacks.
Java's internal security is based on two key ideas:
1. Code that can protect its private state. When the SecurityManager is enabled and a module is sandboxed, it isn't allowed to use reflection to override field or method visibility.
2. Stack walks.
Let's tackle these backwards. Imagine it's time to do something privileged, like open a file. The module containing the file API will be highly privileged as it must be able to access native code. It will have a method called read() or something like that. Inside that method the code will create a new permission object that represents the permission to open files under a certain path. Then it will use AccessController, like this:
FilePermission perm = new FilePermission("/temp/testFile", "read");
AccessController.checkPermission(perm);
The checkPermission call will then do a stack walk to identify the defining module of every method on the stack. Each module has its own set of granted permissions, the access controller will intersect them to determine what permissions the calling code should have. Note: intersection. That means if any unprivileged code is on the stack at all the access check fails and checkPermission will throw an exception. For example, if an unprivileged module registers a callback from a highly privileged module, that doesn't work: the low privileged module will be on the stack and so the privilege is dropped.Access control contexts are themselves reified as objects, so instead of doing a permission check immediately you can 'snapshot' the permissions available at a certain point and use it later from somewhere else. And, starting a thread copies the permissions available at that point into the new thread context. So you cannot, in the simple case, elevate privilege.
Stack walking and permission intersection is slow. It was optimised a lot in Java 9 and 10 so the performance impact of enabling sandboxing is much less than it once was, but it's clearly not zero overhead. Therefore the JVM provides other techniques. One is the notion of a capability, known from many other systems. Instead of doing a permission check on every single file read (slow), do it once and then create a File object. The File object allows reading of the underlying native file via its private fields. Whoever has a pointer to the File object can thus read from it. Because pointers cannot be forged in Java, this is secure as long as you don't accidentally lose your pointer or pass it to code that shouldn't have it.
Sometimes you need to wrap a privileged operation to "dilute" it somehow. For example, imagine you have a module that allows arbitrary socket access. You also have an HTTP client. You would like the HTTP client to have network access, but for it to be usable by other modules that should only be able to contact specific hosts. Given what I've described so far that wouldn't work: the highly privileged code that can do native calls would do a stack walk, discover the unprivileged module on the stack and throw an exception. But there's a fix: AccessController.doPrivileged. This is kind of like sudo. It takes a lambda and truncates the stack that's examined for access checks at the point of use. Therefore it allows a module to use its own assigned permissions regardless of who is calling it. Of course, that is powerful and must be used carefully. In this case the HTTP client would itself check a different, HTTP specific permission. If that permission passed, then it would assert its own power to make arbitrary network connections and go ahead and use the lower level API.
There are a few more pieces but they aren't core. One is the class called SecurityManager. This is the most famous part of the API but in fact, it's no longer really needed. SecurityManager simply delegates to AccessController now. Its API is slightly more convenient for the set of built in permissions. For the purposes of understanding the design you can effectively ignore it. The SecurityManager needs to be activated using a system property as otherwise, for performance reasons, permission checks are skipped entirely at the check sites. Beyond that it can be left alone, or alternatively, customised to implement some unusual security policy. Another piece is the policy language. Permissions are not intrinsic properties of a module in the JVM but rather assigned via an external file. The final piece is the module system. This isn't relevant to the sandbox directly, but it makes it easier to write secure code by adding another layer of protection around code to stop it being accessed by stuff that shouldn't have access to it. After a careful review of the old JVM sandbox escapes from the applet days, the Java team concluded that the module system would have blocked around half of them.
So as you can see the design is very flexible. There's really nothing else like it out there, except maybe .NET CAS but I believe they got rid of that.
Unfortunately there are some pieces missing, if we want to re-awaken this kraken.
The first is that modules have no way to advertise what permissions they need to operate. That has to be specified in an external, per-JVM file, and there are no conventions for exposing this, therefore build tools can't show you permissions or integrate the granting of them.
The second is that some code isn't sandbox compatible. The most common reason for this is that it wants to reflectively break into JVM internals, for example to get better performance. Of course that's not allowed inside a sandbox.
A third is that some code isn't secure when sandboxed because it will, for example, create a File object for your entire home directory and then put it into a global public static field i.e. it doesn't treat its capabilities with care. The module system can help with this because it can make global variables less global, but it's still not ideal.
The final piece is some sort of community consensus that sandboxing matters. Bug reports about sandboxing will today mostly be ignored or closed, because developers don't understand how to use it and don't see the benefit. It's fixable with some better tutorials, better APIs, better tooling etc. But first people have to decide that supply chain attacks are a new thing that matters and can't be ignored any longer.
Not sure if this is something Java enables, but in principle you could do this in a capability-style way as well. Let's say you have an HTTP client module that you want to allow another module to use, but only to make requests to a specific host. You could write a wrapper with a subset of the HTTP client's functionality, only including (for example) a send() method that would send an HTTP request to the specified host. You'd then pass that to the module that you want to be able to make HTTP connections (rather than the raw HTTP client), and provided your wrapper object doesn't expose functionality from the underlying module that would let a client specify arbitrary hosts, you're in a pretty good spot.
Sure, something needs to have permission to use the higher level of privilege. On your typical POSIX OS, your program is probably born with the ability to create arbitrary TCP/UDP sockets by default; on a capability OS, maybe you've explicitly provided it with access to your network stack. Regardless, at the entry point to your program you presumably have modules providing arbitrary network access in scope somehow.
If I'm understanding correctly, the case you described is that you have an HTTP client module that you'd like to have direct access to the network, but you'd like to restrict the consumers of the HTTP client to only querying certain hosts. From the start of your program, you'd instantiate an HTTP client (passing it a capability to use the network interface) then instantiate one of those HTTP client proxy objects that only allows communication with one host (passing it a capability to use the HTTP client). From there, you pass the capability to that proxy object to the unprivileged consumer of the module.
This seems to work without any kind of stack walking authentication logic, just normal variable scope, provided the language is capability-based. Am I missing something?
Indeed, if you look at the history of Java sandbox escapes they are largely confused deputy attacks: some privileged code source can be tricked into doing something it shouldn’t do.
You're imagining a system where there's no specific authentication system for code. Instead in order to use a library, you need to explicitly and manually obtain all the capabilities it needs then pass them in, and in main() you get a kind of god object that can do everything that then needs to be progressively wrapped. If a library needs access to a remote service, you have to open the socket yourself and pass that in, and the library then needs to plumb it through the whole stack manually to the point where it's needed. If the library develops a need for a new permission then the API must change and again, the whole thing has to be manually plumbed through. This is unworkable when you don't control all the code in question and thus can't change your APIs, and as sandboxing is often used for plugins, well, that's a common problem.
There's no obvious way to modularise or abstract away that code. It can't come from the library itself because that's what you're trying to sandbox. So you have to wire up the library to the capabilities yourself. In some cases this would be extremely painful. What if the library in question is actually a networking library like Netty? There could be dozens or hundreds of entry points that eventually want to open a network connection of some sort.
What does this god object look like? It would need to hold basically the entire operating system interface via a single access point. That's not ideal. In particular, loading native code would need to also be a capability, which means any library that optimised by introducing a C version of something would need to change its entire API, potentially in many places. This sort of design pattern would also encourage/force every library to have a similar "demi-god" object approach, to reduce the pain of repeatedly passing in or creating capabilities. Sometimes that would work OK, other times it wouldn't.
The stack walking approach is a bit like SELinux. It allows for a conventional OO class library, without the need for some sort of master or god object, and all the permissions things need can be centralised in one place. Changes to permissions are just one or two extra lines in the security config file rather than a potentially large set of code diffs.
Now all that said, reasonable people can totally disagree about all of this. The JVM has been introducing more capability objects with time. For example the newer MethodHandle reflection object is a capability. FileChannel is a capability (I think!). You could build a pure capability language that runs on the JVM and maybe someone should. Perhaps the usability issues are not as big a deal as they seem. It would require libraries to be wrapped and their APIs changed, including the Java standard library, but the existing functionality could all be reused. The new libraries would just be a thin set of wrappers and forwarders over pre-existing functionality, but there'd be no way for anything except the god object to reach code that'd do a stack walk. Then the security manager can be disabled, and no checks will occur. It'd be a pure object capability approach.
You don't need to do this. There are a variety of ways to handle this, just as you would any other kind of dependency injection:
1. Design libraries to actually be modular so that dependencies (including capabilities) can be injected just where they are needed.
2. Pass in a factory object that lets the library construct sockets as and when it needs them. You can then enforce any arbitrary checks at the point of creating the socket. (This is much more flexible than a Java policy file).
3. Use a powerbox pattern [1] to allow the user to be directly asked each time the library attempts to open a socket. This is not always good UX, but sometimes it is the right solution.
> If the library develops a need for a new permission then the API must change and again, the whole thing has to be manually plumbed through.
Capturing permission requirements in the API is a good thing! With the stack walking/policy based approach I won't know the library needs a new permission until some library call suddenly fails at runtime.
OK, so in a pure capability language how would you implement this: program A depends on dynamically loaded/installed plugin B written by some third party, that in turn depends on library C. One day library C gets a native implementation of some algorithm to speed it up. To load that native library requires a capability, as native code can break the sandbox. However:
1. You can't change the API of C because plugin B depends on it and would break.
2. You can't pass in a "load native library" capability to plugin B because you don't know in advance that B wants to use C, and if you did, B could just grab the capability before it gets passed to C and abuse it. So you need to pass the capability directly from A to C. But now A has to have a direct dependency on C and initialise it even if it's not otherwise being used by A or B.
Stack walking solves both these problems. You can increase the set of permissions required by library C without changing its callers, and you don't have the problem of needing to short-circuit everything and create a screwed up dependency graph.
With the stack walking/policy based approach I won't know the library needs a new permission until some library call suddenly fails at runtime
You often wouldn't need to. What permissions a module has is dependent on its implementation. It's legitimate for a library to be upgraded such that it needs newer permissions but that fact is encapsulated and abstracted away - just like if it needed a newer Java or a newer transitive dependency.
Now, I'm a little outside my area of expertise due to not having worked with capability systems very much yet. (There aren't that many of them and they're still often obscure, so even just trying to gain experience with them is difficult at this point.)
But that said... in an ideal capability system, isn't the idea that native code could just break the sandbox also wrong? I would imagine that in such a system, depending on another module that's running native code would be just fine, and the capability system's constraints would still apply. Maybe that could be supported by the OS itself on a capability OS; maybe the closest thing we'll get to native code for that on our existing POSIX systems is something like WASI[0].
> You often wouldn't need to. What permissions a module has is dependent on its implementation. It's legitimate for a library to be upgraded such that it needs newer permissions but that fact is encapsulated and abstracted away - just like if it needed a newer Java or a newer transitive dependency.
If our goal is to know that the dependencies we're using don't have more authority than they need, isn't it a problem if a module's permissions may increase without explicit input from the module's user (transitive or otherwise)?
[0] https://hacks.mozilla.org/2019/03/standardizing-wasi-a-webas...
> isn’t it a problem if a module’s permissions may increase without explicit input from the module’s user (transitive or otherwise)?
Exactly right.
[1]: https://www.cl.cam.ac.uk/research/security/ctsrd/cheri/
The modules permissions can't increase without explicit input e.g. changes to the policy file. But the person who cares about the sandbox integrity is the user of the overall software or computing system. The plugin developer doesn't really care how the API is implemented or what permissions it needs. They just want it to work. The person who cares is the person who owns the resource or data an attacker may be trying to breach.
The Java approach is nice because it avoids any need for DI. DI is not a widely accepted pattern. There are no DI engines that would have any support for this kind of policy-driven injection. And whilst popular in some areas of software like Java web servers, it hardly features in most other languages and areas, there are no programming languages with built in support for it and that includes modern languages like Kotlin. DI engines meanwhile have changed how they work pretty radically over time - compare something like the original Spring XML DI to Guice to Dagger3. Plus, DI is awkward when the dependency you need isn't a singleton. How would I express for example "I need a capability injected that gives me access to the $HOME/.local/cache/app-name directory"? Annotation based DI struggles with this, but with the AccessController it's natural: the component just requests what it needs, and that's checked against a policy, which can be dynamically loaded from a file, or built by code.
You are confusing DI frameworks (often terrible) with the general concept of dependency injection, which is in fact extremely widely used.
The problem with Java’s stack walking is that it is too complex and too easy to find privileged code that can be coaxed into performing unintended operations. There are plenty of old write ups of Java sandbox bypass bugs due to this, eg http://benmmurphy.github.io/blog/2015/10/21/zdi-13-075-2013-...
You're right that java.io.File isn't a capability. It just represents a file path with a few utility methods to list files, and therefore does a stack walk when you try to access the filesystem. A FileChannel is a file capability in the sense I meant above, because it represents an opened file, not a path. There's an access check once, when it's opened, and then the rest of the time there aren't any stack walks.
It's a pity that Ben Murphy didn't write up all his bugs. There are only two listed there. A few patterns cropped up repeatedly in the old sandbox escapes:
1. Accessing internal code you weren't meant to have access to. Often, some sorts of privileged pseudo-reflection API. Fixing this is the goal of Jigsaw.
2. Serialization acting as a back door, allowing the internal private state of classes to be tampered with. Serialization security has been improved with time and they're now working on a feature that will make it harder to screw this up, by allowing serialization to use normal constructors instead of this ad-hoc form of reflection.
3. Overly general frameworks that allowed attackers to construct nearly arbitrary programs out of privileged objects by chaining them together (this crops up in gadget attacks too). There's probably no platform level fix for this, people just have to be aware of the risks when working in sandboxed context.
I don't think a pure capability language is workable, to be honest. At least not at a large scale. In the purest sense you need a god object passed into the start of your program which vends all possible capabilities, and every library would require the user to construct and pass in all resources it needs externally, including things it might possibly need. And that code isn't easily modularised because you can't just use helpers provided by the library itself: that's the very same code you're trying to sandbox. There's no good way to make that usable or extensible. The combination of stack walking and capabilities lets you find the right balance in terms of API design between simplicity of use and sandbox simplicity.
* The (now defunct) E programming language: http://erights.org and Joule: https://en.wikipedia.org/wiki/Joule_(programming_language) * Pony: https://www.ponylang.io * Monte: http://www.monte-language.org * The Midori project at Microsoft: http://joeduffyblog.com/2015/11/10/objects-as-secure-capabil...
It's actually not at all unworkable to use object-capability for large programs. In fact, one of the main benefits of ocaps is how well it aligns with well-established good software design principles such as dependency injection, avoiding singletons, avoiding global mutable state, and so on.
I'd like to see a more modern attempt that wasn't as totally obscure as those other languages. However, nobody is doing that.