We have a suite of applications that depend on the sharing of a directory/files on external storage.
I've currently opted out of the Android 10 OS changes to scoping (requestLegacyExternalStorage), but this is going away and I've spent many hours trying to find a solution for simply sharing files between applications.
The only solutions that I see offered are:
SAF - which appears to make the user choose through UI. This is completely undesirable.
Use a File Content Provider - the way I understand this, I would have to make the user install an apk with my provider in it before installing any of my applications. Forcing the user to install two apks to run one application is very undesirable. (Yes, they could both be in one apk manifest but who knows which of my suite they will want to install)
Media Store - My understanding is that this also forces the user to pick something he should have no knowledge of - and is really intended for audio, video, image and downloaded directory.
Am I missing a solution for these simple requirements?
Am I missing a solution for these simple requirements?
There is no simple solution. You would basically need to have each app have its own copy of the shared data (to deal with potential uninstalls) and have some sort of synchronization protocol so each app in the suite can inform others about changes to their copy of the data.
Using SAF is the simplest approach for your scenario. Or, move the data off the device into "the cloud".
My understanding is that this also forces the user to pick something he should have no knowledge of
It is the user's device. It is the user's storage. If you put files in a user-visible location on the user's storage, they are the user's files. Your apps are merely one set of tools for working with those files, nothing more.
Assume we have a process that may dlopen() some third-party library. This library may perform open("write_only_logfile", O_WRONLY) on some file to which user has only write access. We need to have an ability to be notified if this library attempts to open a file, so later we may dup() returned descriptor and redirect output.
There are few restrictions that make interception harder:
LD_PRELOAD is forbidden - no way to hook open()
inotify(7) doesn't help because user has no read permissions on "write_only_logfile" and it is owned by admin
we have no access to library sources and therefore cannot modify it
"write_only_logfile" is hardcoded inside the library, so we cannot pass another name to perform redirecting
I'm wondering if Linux has an efficient way to help in such situation.
Especially taking in account the fact that process may open() miscellaneous files pretty often.
P.S. To avoid confusion and understand better - it is a regular Android application with loaded JVM. If app hangs (so called ANR) - system sends SIGQUIT to it. Signal is received via dedicated thread that open()s /data/anr/traces.txt and writes JVM state to it. These data extremely useful for debugging. But app cannot read that file directly because of security reasons (All applications write to it, so there may be somewhat sensitive). Anyway I believe that it is absolutely fair to intercept content that my process would write to it.
P.S.S. In the worst case it is possible to find JVM library image (libart.so) and manually patch jump slot for open(). But it doesn't sound well.
Sounds like you are in troublesome situation. Most solutions briefly mentioned below are guaranteed to interfere with SELinux, so don't take my word for any of that.
Debugging your own process with strace to intercept open is one of usual solutions on normal Linux. I am not sure if it would work in Android; it certainly might become off-limit for non-debuggable apps starting in some new versions (if it is has not been banned yet).
seccomp-bpf is another possibility. Might not be available on older Android versions, but since Android O seccomp is going to be a guaranteed part of Android security getup. Intercept open in warn-only mode and give control back to yourself when something interesting happen (via debugging or signals).
If /data/anr/traces.txt is opened on-demand, you should be able to observe that by watching contents of /proc/self/fd/ with inotify or via polling. You might be able to reduce impact of races by setting io niceness of the opening thread…
All of above are only partial solutions, you still might need to decode actual open syscall that happened (strace source code might be helpful there for strace/seccomp solutions, readlink for /proc/self/fd/) and act upon it (dup2, as you already mentioned).
"write_only_logfile" is hardcoded inside the library
Is it possible to modify the memory of data segment of the library/executable? Afaik mprotect and PROTECT_EXEC in particular have been heavily restricted, but at least mmap is certainly permitted (to support JIT compilers etc). It might be possible to cook something up to edit the string constant in place (as long as doing so is possible and allowed, I am not sure myself about that).
If this is just about redirecting writes (and reads) to a single file, you could run the application in a mount namespace with a suitable bind mount for that particular file. Setting things up in this way probably requires a small SUID binary.
A more general solution quickly approaches a union file system, and getting it right is quite hard. Even the in-kernel union file system, overlayfs, does not manage to provide full POSIX semantics.
You need LD_PRELOAD to hook an application. To hook a third-party library, just load your hook normally before the library (or have it in your executable).
Assuming the library calls open from libc and not the corresponding syscall directly, and that it is linked in a normal way, you just have a function named open somewhere in your code. Make it call open from libc (RTLD_NEXT or whatever). The third-party library (and all other libraries of course) will resolve its open symbol to your function.
I found an android app named Super Erase that deletes files and folder permanently from android device so that the file deleted cant be recovered anymore..here is the application i am talking about ...but i was wondering how to that and i know it is made with android studio ..i tried the regular way to delete file.delete() but still the file can be recovered.can i have any help .
For starters, secure file deletion on flash media is a complex problem, with no quick and easy answers. The paper Reliably Erasing Data From Flash-Based Solid State Drives gives a good overview of the problems, the potential solutions, and their limitations. They conclude that
For sanitizing entire disks, ... software techniques work most, but not
all, of the time. We found that none of the available software
techniques for sanitizing individual files were effective. [emphasis added]
NIST 800-88 also has a good overview of the technology trends contributing to the problem, along with some minimum recommendations (appendix A) for Android devices. However they tend to be either whole-disk erasure (factory reset), or rely on cryptographic erasure (CE), rather than being general file erasure methods.
But all is not lost. Even if you can't sanitize individual files, you could hope to wipe all the unallocated space after deleting files. The article Secure Deletion on Log-structured File Systems (Reardon, et al.) describes a fairly promising way to do that in user-mode software. Android's internal memory uses (always?) a log-structured file system.
This paper's "purging" method does not require kernel-level access, and doesn't seem to require any native code on Android. (Note that the term "purging" is used a little differently in documents like NIST 800-88.) The basic idea is to delete all the sensitive data, then fill in the remaining space on the drive with a junk data file, and finally delete the junk data file.
While that takes more time and effort than just overwriting the deleted files themselves (several times in different patterns), it seems to be very robust even when you have to deal with the possibility of wear-leveling and log-structure FS.
Caveat and Further Measures
The main caveat for me is about the conditions mentioned by Reardon et al. in the above paper:
Purging will work for any log-structured file system provided both the
user’s disk quota is unlimited and the file system always performs
garbage collection to reclaim even a single chunk of memory before
declaring that the drive is unwritable. [emphasis mine]
The second condition seems pretty likely to be fulfilled, but I don't know about the first one. Does Android (or some manufacturers' versions of it) enforce quotas on disk space used by apps? I have not found any info about user quotas, but there are quotas for other niches like browser persistent storage. Does Android reserve some space for system use, or for each app's caching, for example, that can't be used for other things? If so, it should help (albeit with no guarantees) if we begin purging immediately after the sensitive files are deleted, so there is little time for other filesystem activity to stake a claim to the recently freed space.
Maybe we could mitigate these risks by cyclical purging:
Determine the remaining space available (call it S) on the relevant partition, e.g. using File.getUsableSpace()
Write a series of files to the partition; each one is, say, 20% of the initial S (subject to file size limits).
When we run out of space, delete the first couple of files that we created, then write another file or two as space allows.
Repeat that last step a few times, until you've reached a threshold you're satisfied with. Maybe up to the point where you've written 2*S worth of filler files; tweak that number to balance speed against thoroughness. How much you actually need to do this would be an area for more research.
Delete the remaining filler files.
The idea with cyclical purging is that if we run out of quota to overwrite all free space, deleting the filler files just written will free up more quota; and then the way log-structured filesystems allocate new blocks should allow us to continue overwriting the remaining blocks of free space in sequence, rather than rewriting the same space again.
I'm implementing this method in a test app, and will post it when it's working.
What about FAT-formatted microSD cards?
Would the same methods work on external storage or microSD cards? FAT is block-structured, so would the purge method apply to FAT-formatted SD cards?
On most contemporary flash memory devices, such as CompactFlash and
Secure Digital cards, [wear leveling] techniques are implemented in
hardware by a built-in microcontroller. On such devices, wear leveling
is transparent and most conventional file systems can be used on them
as-is. (https://en.wikipedia.org/wiki/Wear_leveling)
...which suggests to me that even on a FAT-formatted SD card, wear leveling means that the traditional Gutmann methods would not work (see his "Even Further Epilogue") and that a method like "purging" would be necessary.
Whether purging is sufficient, depends on your security parameters. Wear leveling seems to imply that a block could potentially be "retired" at any time, in which case there is no way to erase it without bypassing the microcontroller's wear leveling. AFAIK this can't be done in software, even if you had kernel privileges; you'd have to design special hardware.
However, "retiring" a bad block should be a fairly rare event relative to the life of the media, so for many scenarios, a purging method would be secure enough.
Erasing the traces
Note that Gutmann's method has an important strength, namely, to erase possible traces of old data on the storage media that could remain even after a block is overwritten with new data. These traces could theoretically be read by a determined attacker with lots of resources. A truly thorough approach to secure deletion would augment a method like Gutmann's with purging, rather than replacing it.
However, on log-structured and wear-leveled filesystems, the much bigger problem is trying to ensure that the sensitive blocks get overwritten at all.
Do existing apps use these methods?
I don't have any inside information about apps in the app store, but looking at reviews for apps like iShredder would suggest that at best, they use methods like Reardon's "purging." For example, they can take several hours to do a single-pass wipe of 32GB of free space.
Also note limitations: The reviews on some of the secure deletion apps say that in some cases, the "deleted" files were still accessible after running the "secure delete" operation. Of course we take these reviews with a grain of salt -- there is a possibility of user error. Nevertheless, I wouldn't assume these apps are effective, without good testing.
iShredder 4 Enterprise helpfully names some of the algorithms they use, in their app description:
Depending on the edition, the iShredder™ package comes with deletion
algorithms such as DoD 5220.22-M E, US Air Force (AFSSI-5020), US Army
AR380-19, DoD 5220.22-M ECE, BSI/VS-ITR TL-03423 Standard,
BSI-VS-2011, NATO Standard, Gutmann, HMG InfoSec No.5, DoD 5220.22 SSD
and others.
This impressive-sounding list gives us some pointers for further research. It's not clear how these methods are used -- singly or in combination -- and in particular whether any of them are represented as being effective on their own. We know that Gutmann's method would not be. Similarly, DoD 5220.22-M, AFSSI-5020, AR380-19, and Infosec No. 5 specify Gutmann-like procedures for overwriting sectors on hard drives, which would not be effective for flash-based media. In fact, "The U.S. Department of Defense no longer references DoD 5220.22-M as a method for secure HDD erasure", let alone for flash-based media, so this reference is misleading to the uninformed. (The DoD is said to reference NIST 800-88 instead.) "DoD 5220.22 SSD" sounds promising, but I can't find any informative references for it. I haven't chased down the other algorithms listed, but the results so far are not encouraging.
When you delete file with standard methods like file.delete() or runtime.exec("rm -f my_file") the only job that kernel does is removing info about file from auxiliary filesystem structures. But storage sectors that contain actual data remain untouched. And because of this recovering is possible.
This gives an idea about how we can try to remove file entirely - we should erase all sectors somehow. Easiest approach is to rewrite all file content with random data few times. After each pass we must flush file buffers to ensure that new content is written to storage. All existing methods of secure file removal spin around above principle. For example this one. Note that there is no universal method that works well across all storage types and filesystems. I guess you should experiment by yourself and try to implement some of the existing approaches or design your own. E.g. you can start from next:
Overwrite and flush file 10 times with random data (use FileOutputStream methods). Note!!! Don't use zeros or another low entropy data. Some filesystems may optimize such sparse files and leave some sectors with original content. You can use /dev/urandom file as source of random data (this is a virtual file and it is endless). It gives better results and works faster then well-known Random class.
Rename and move file 10 times. Choose new file names randomly.
Then truncate file with FileChannel.truncate().
And finally remove file with File.delete().
Of course you can write all logic in native code, it may be even somewhat easier than in Java. Described algorithm is just an example. Try doing in that way.
The standard filesystem API won't give you a simple function call for that.
You will have to use the underlaying native API for FileIO. Although I have never used it, theres a library for that:
https://github.com/johanneslumpe/react-native-fs
There are two answers to this question.
First, to answer the direct question of how some of these apps might be doing secure single file delete: what you do is actually open the file and replace the contents with zeros many times. The method sounds stupid, but I have worked with filesystem-level encryption on Android in the past and I found that the above holds true for many secure file delete solutions out there. For a seemingly compliant security, you can repeat writing zeros 7 times (or whatever the NIST standards specify for your hardware type).
Charset charset = StandardCharsets.UTF_8;
String content = new String(Files.readAllBytes(path), charset);
content = content.replaceAll("*", "0");
Files.write(path, content.getBytes(charset));
The right answer to this question is however different. On modern SSD drives and operating systems, it is insecure to delete single files. Therefore, these apps don't really offer a compelling product. Modern operating systems store fragments of the file in different places, and it is possible that even after you have zeroed out the most recent version of the file block-by-block and also overwrote all metadata, that a fragment from an older version of the file might be left over in another part of the drive.
The only secure way to delete sensitive content from a disk is to zero out the entire disk multiple times before discarding the disk.
#LarsH's answer about wiping all unallocated space after deleting files is compelling, but perhaps impractical. If you simply want to secure delete files so no one can scan the disk to recover it, then a better solution is the full-disk encryption. This was in-fact the entire appeal of full-disk encryption. This is why Apple stopped supporting secure file delete in their Mac OSX and iOS, and switched to full-disk encryption as default on all iPhones. Android phones have full-disk encryption as well now.
EDIT:
If you are looking for a true solution for a customer, your best bet is to use single file encryption. Once you destroy your key which only your app would know, there is no way to decrypt the file even if someone found it on the disk.
There exists no real solution for deleting files securely on SSDs. You can only give a false sense of security to non-technical people who still remember the old HDD days.
How can we make Android assets secure so that no one can read them after app deployment?
There is nothing you can do that will stop a determined attacker from reading them.
Using your own application level encryption would at least make the problem unique to your application, but someone could still do code analysis of your app to figure out how to decrypt them.
The platform's limited copy protection mechanisms are weaker, because they only have to be defeated once for all applications (such as by rooting the phone).
Assets utilized by platform functionality would also vulnerable to a modified platform configured to dump out copies of them.
Do what is easy to stop casual copying by unsophisticated users if you like, but then save your time and energy for battles you can actually win, such as the quality of your application.
I'm porting a rather large game engine written in C++ from Windows/Mac to Android. There is a lot of pre-existing code to read assets for games. In addition, there is quite a bit of code doing file system calls (stat'ing the files to make sure they exist, looking up all of the files and directories inside of a directory, etc.)
Right now, I'm focusing on just getting something up and running as quickly as possible, so I'd prefer not to have to rewrite a lot of this. What would be a good way of getting our game assets onto the device and accessing them with minimal changes to our existing standard C++ file system API usage?
I've got some basic support implemented already using the Asset Manager API, but that doesn't support the file system calls and I'm concerned that the 1 MB asset size limit is going to bite me at some point.
I've also looked at OBB, but the tools for creating an OBB file don't look like they are part of the current SDK/NDK. Otherwise, that looks like it would be perfect.
Is it a horrible idea to package up all of the files and just extract them on the SD Card the first time the app is run? Or is there some better way of dealing with this?
Update: I'm also not very concerned on being able to run on a broad range of devices, I am specifically looking at newish tablets, probably the 10.1" Samsung Galaxy tab.
We ran into a similar problem in developing our (data-file-heavy) app, and we ended up deciding to keep the APK tiny and simply download our data files on first run; they're going to have to be downloaded either way, but a small APK works much better on older devices without a lot of internal storage. Plus, you can potentially rig up a way for people to copy over the files directly from their computer if they have a limited data plan or a slow internet connection on their phone.
The "Downloader" sample app in apps-for-android (confusingly buried under "Samples") is almost a fully-implemented solution for this - you can pretty much just plug in the particulars of your data files and let it do the rest.
I wrote an app that relies on putting a good amount of native code into the Android filesystem. I did this by packaging the files into the APK as 'resources'. Instead of pushing them to the SD card, you can put then into the application's private namespace, I.E. /data/data/com.yourdomain.yourapp/nativeFolder.
For details on how to accomplish this, you can see my answer to this question.
It's fairly simple to package to just unpack them on the first run and never worry about them again. Also, since they're under the application's namespace, they should be deleted if/when someone were to decide to delete your app.
EDIT:
This method can be used to put anything into the app's private area; /data/data/com.yourdomain.yourapp/
However, as far as I know, your application has to be the one to create all the folders and sub-folders in this area. Luckily this is fairly easy to do. For example to have your app make a folder:
Process mkdir = Runtime.getRuntime().exec("mkdir " +localPath);
That works as it would in most linux shells. I walked through the assets folder I packaged into my APK, made the corresponding directories and copied all the native files to those directories.
What you might be more concerned with is the limited Android shell. There are many commands that you might want that aren't present. stat for example isn't available, so all of this may be moot if your native code can't make it's system calls.