I came across this article while surfing the internet. I wanted to share this with you guys, and see what your feelings were on this.
"Mobile Device Security and Android File Disclosure
Back in November, Thomas Cannon brought to light an issue within the Android operating system. Specifically, he found that it was possible to obtain the contents of files on an Android device by simply persuading its owner to visit a web site under attacker control. The issue only garners a 3.5 CVSS score, but yet it’s still fairly serious.
Thomas reported this issue responsibly to Google and they took it seriously. However, since then they have come back with a ridiculous remediation plan. Granted, its probably not entirely Google’s fault, but the overall situation looks very bleak for Android.
The problem is that Google stated that a fix will be available as part of an update to the upcoming Android 2.3. While that, in itself, may not be totally ridiculous, the reality of the situation is that Google is only one party involved in Android. There are two other groups, namely OEMs and Carriers, that must also do their part in getting the fix to users. Although Android devices are becoming increasingly functional, the security posture remains abysmal.
The security posture for desktop applications has improved vastly with all of the sand-boxing, automatic updates, and various other exploit mitigation technologies. Meanwhile, Android includes almost none of existing security protections. In fact, mobile users are being left out in the cold, unable to get a patch for a trivially exploitable cross-zone issue. For that matter, they can’t even control whether their device’s browser automatically downloads files or not.
This situation is not news, rather it is a sad fact. It is totally unfair for end users to be left out to fend for themselves. After all, they are paying a small fortune for these devices and the service to be able to use them. Hopefully the vendors involved will wake up before a network worm outbreak occurs.
Originally, Thomas disclosed the details of his bug on his blog. Later, he removed some details to help protect users. I believe that responsible disclosure is a two-way street that requires responsibility on both sides. Since Google, OEMs, and carriers all continue to act irresponsibly, it is necessary bring more attention to this issue and the situation as a whole.
I spent a little time and managed to recreate the issue with nothing more than HTML and JavaScript. As of today, I have released a Metasploit module to take advantage of the flaw. It is available in the latest copy of our Framework product, or you can view the source via the link to our Redmine project tracker above.
Before I go deeper into the consequence of this bug, I want to point out that Thomas outlined several workarounds for this vulnerability in his blog.
Now, take a deep breath give some thanks to the fact that, under Android, most every process runs under a separate, confined, unix-style user account. This design feature partially mitigates this issue, lowering confidentiality impact to “Partial” and bringing the CVSS score from 5 to 3.5. That said, an attacker can still gain access to some pretty interesting stuff.
For starters, an attacker can steal any world-readable file. In my tests it was possible to get potentially sensitive information from the within the “proc” file system. This type of information could include kernel versions, addresses, or configuration that can be used enhance further attacks.
Also, you can snarf any files that are used by the browser itself. This includes bookmarks, history, and likely more. This kind of information could potentially be embarrassing or possibly even give an attacker access to any saved passwords or session cookies you might have stored.
Perhaps the easiest win though, is that you can grab anything off of the SD card. You might ask, “Anything?! What about the user separation?” Well, because the SD card has been formatted with the “vfat” (aka “fat32”) file system, there is no concept of ownership. All files are owned by the same user id since the file system itself cannot encapsulate who created which file. As Thomas said, files in the SD card that have predictable names are ripe for the picking. This includes pictures and movies. These may in fact be some of the most private data on your device.
In conclusion, I hope that the Android security debacle will get resolved as soon as possible. If Google, OEMs, and carriers can’t work it out, perhaps another party will step in to maintain the operating system. I believe this could be very similar to the way various Linux distributions operate today. If the situation is not resolved, I fear the Android device pool could become a seething cesspool of malicious code..."
Here is the address
http://blog.metasploit.com/2011/01/mobile-device-security-and-android-file.html
Sent from my PC36100 using XDA App
Shocking. Thanks for the info.
Nice find. You are right that oems and manufactures need to stay on top to mantain security. Hopefully meaningful post like this will make users aware of the possible dangers of the internet, data, and phone usage
Sent from my ADR6300 using Tapatalk
Ouch. Wish Android updates were like iOS..
Android is open, one of the main assumptions is that there is no single company, which controls it. I could create my own phone with Android, sell it to people and give them no support at all - Google can't do anything about it.
There is only one solution to this problem: people have to choose their phones wisely. People look at phone specs, at CPU, RAM, camera, but they ignore future support and openess. Recently Motorola has stated they will lock bootloaders in their future phones. People will go for these phones anyway and then they will complain they can't do anything with some horrible bugs, they will complain about Android and Google, but they should complain about Motorola and themselves. While Nexus S owners will have same bugs fixed by both Google and community.
Choose your phones wisely.
SD with vfat...good catch. Horrible bug while many users trying to move their apps to SD. And maybe 80-90% of the apps in the market require modify SD card perm? Horrible. Verizon SGS is screwed since that phone have little internal and lots of external SD.
I'm so glad you guys came across this thread, and it didn't get lost in all the other threads. I hope some of the devs see it. Can a fix be implemented at the Rom or kernal level?
Sent from my PC36100 using XDA App
Cydia Substrate is a code modification platform. It can modify the code of any major process, whether that code is written in Java or C/C++. It has been designed to support an ecosystem where many developers are interested in hooking the same processes. It is designed to be powerful and efficient.
== How do I get Substrate? ==
You can either download it from the Play Store or directly from its website.
== How do I develop for Substrate? ==
You download its SDK using the Android SDK manager or from its website. There is extensive documentation on the website.
== What is Substrate's website? ==
http://www.cydiasubstrate.com/
== How is this different from Xposed? ==
Many compare it with Xposed, but Xposed only supports a single use case: hooking Java functions inside of app_process. Substrate can hook native code, such as is required to modify the way styles are loaded inside of the Android asset manager. There are many other differences, however, as Substrate's API is based on five years of experience managing a community of runtime code modification for iOS. I normally avoid doing direct comparisons, but after attending Big Android BBQ and presenting on Substrate, I have been encouraged to make the differences and advantages of Substrate's approach more explicit here on XDA.
Xposed requires an inverted form of logic based on "before" or "after" hooks while Substrate lets the developer use more straightforward "replace and call previous" implementations. This also enables more complex interactions with the previous implementation that have been shown to be valuable among the thousands of developers using Substrate on other platforms. Xposed attempts to offer something similar with "replace" hooks, but those do not provide access to the previous implementation, and while Xposed provides a way to call the "original" implementation, that skips any other hooks that might be stacked.
Xposed requires the developer to find a safe moment to interact with the class being hooked. To make this possible, there are numerous lifecycle events such as "VM loaded", "package loaded", and "command line application started". However, this does not solve the problem that touching classes can change the order in which they statically initialize. This also means that it will not be possible to provide declarative syntax wrappers (such as Logos, which developers use on iOS) on top of Xposed, as this context will have to be made implicit in imperative logic. Substrate solves this class initialization problem by allowing developers to hook the classloader itself, getting a callback when a class is "linked" so that the developer can find a class loaded in any classloader (even as a plugin to an application an hour after that application starts, where the code is downloaded as a .dex from the Internet).
Xposed has a method hook implementation that makes it lose track of which method was hooked, requiring it to do a lookup every time such a method is called. This implementation is currently linear in the number of hooks, making it slow down the more hooks you install. Worse, there is a high constant multiplier on this algorithm, as the comparison between entries is very expensive (and was made more expensive when recently fixing a longstanding bug caused by this lookup being slightly incorrect). Substrate, in comparison, uses runtime code generation to avoid the need to every look anything up at runtime: you can use Substrate to hook small functions in tight loops without experiencing the same kind of performance issues you would see with Xposed.
Substrate is also designed with a different user focus: while it currently has a setup interface, it would prefer to not have any UI at all (and this will be strived for in subsequent versions, assuming anyone cares to use it). Upgrades to Substrate can be automatically installed by the Play Store and do not require the user to interact with Substrate for the changes to "stick". Substrate itself is distributed via Play. Rather than confine these kinds of modifications to advanced users who use forums such as XDA, the idea is that everyone should have access to using this kind of technology. If you have a ROM or another store in which you'd like to see Substrate distributed, I would be more than happy to talk to you about this to make that happen, and these installations will be fully supported.
For some more information on the differences between Xposed and Substrate (or if you are wondering why you should bother paying any attention to things that I say, as maybe you don't remember me from my earlier Android projects), I encourage you to read the comments I left a couple posts down from here on this thread that describe the history of Substrate, how I fit into the Android ecosystem, and more about how Substrate differs from Xposed. I will also likely be posting the talk I gave at Big Android BBQ (with either notes to go along with each slide or in the form of a video I will record re-giving the talk and advancing the slides), which might make some of these things more clear.
Current Changelog
[this is the changelog from Play, which has been compressed slightly. I will bring the more full changelog back, as I have it saved somewhere, and put it here or link it here]
v0.9.4011:
* fix decoder bug inside ARM emulator
* support Genymotion Intel emulator
* add symbol names for Moto X
v0.9.4010: critical Android 4.3 fix, avoid old Superuser bug
^^ must install before Android 4.3 OTA!
v0.9.4009: work around Xposed bug, 4.2 fix, better errors
v0.9.4008: HTC linker path patch, limit symbol exports
v0.9.4007: RAZR i 4.1.2, detect HTC override, avoid ps
v0.9.4005: incompatibility detector, avoid mount/ln/mkdir
v0.9.4004: Holo, Script Failure, detect physical /vendor
Comments from Developer
So, yeah: I'm the developer of Cydia Substrate, the framework everyone uses on iOS to do runtime code modification. Back in 2011, I gave a talk at Android Open along with a demo of Substrate running on Android 3.0. However, after some in-depth discussions with people there who were interested, I realized that what I had at the time "wasn't sufficient": it was just the core of an implementation, not an end-to-end offering. By the time it had everything I felt it needed to launch--including a comprehensive website filled with documentation, a configuration application to install with it, fully tested support for both ARM and x86, a forward-compatible pure Java API vetted by a bunch of the top people in the iOS modding community (as I feel like breaking APIs after launch is one of the more evil things a framework can do), and an extension that would make sense to end users that they could try (so that trade press wouldn't be horribly confused, as I knew they would report on the release)--it was already 2013.
http://www.youtube.com/watch?v=tA9cnemnQ0A <- Android Open 2011 keynote teaser
As many people then know, I released it in June. A lot of people have tried it (165k installs just from Play, and another 20k downloads of the APK off the site), and many of those people even like it enough to keep it installed and leave positive reviews, despite there being almost nothing available to use with it except WinterBoard (which I really only did as a demonstration). However, I also get comments from people who seem to believe I'm some kind of "interloper" in the world of Android. Additionally, there are the people who leave reviews saying stuff like "this is stupid, we already have Xposed" (sometimes then explicitly adding in the "go home to iOS" kind of spiel). The #1 complaint, however, is "nothing I can do with it", because developers never seem to talk about it or use it much, and the people installing it are all end users. Clearly this isn't the kind of reaction that I thought would happen, especially after having discussed Substrate at length with pulser_g2 before launch (who said that the community here tends to be very good about judging things on their usefulness and technical merits as opposed to having emotional attachments).
http://www.cydiasubstrate.com/ <- Cydia Substrate
Given this, and after an encouraging back/forth I had with some people on reddit's Android subreddit a few days ago in some threads about the analysis I did of that recent Android iMessage client (people who didn't know much about the ways in which Substrate is very different than Xposed in capability and focus), I figured I'd finally make a post on XDA. I kind of had been waiting to make this post as well, honestly (as again: I like things to be more perfect before I release them than maybe people are used to around here ;P), but it seems like I'm now waiting for something that is itself causing the delay (I had really expected to do this in July, before the whole thing got more actively depressing). This is clearly that post ;P. I've responded to a bunch of other threads here talking about Substrate (and the many other Android projects I've released) in the past, but this is the first time I've actually started a thread.
(In specific, Substrate currently doesn't support some Samsung devices due to a change they make to the linker paths, and I wanted to have 100% device coverage before making the inaugural XDA post. However, I'm finding it very demotivating to spend the time to think through all the options I've been considering for workarounds given the overall lackluster reaction to my work, so I'm not even making fast progress anymore: I tend to work on the things that people react positively to, and while I got a lot of positive reactions on the balance from users, I got much less than I expected from developers given how many people use Substrate on iOS and how powerful the framework is. I think, from some conversations I've had, this is largely due to confusion over how Substrate on Android relates to Xposed, which many people seem to think of as the "home-town competitor" "that does the same thing". I thereby figure that I may as well attempt to directly address that core motivation problem, to see if I should even bother continuing spending time helping out in this community, hence this ludicrously long and highly personal post about what is essentially a technical framework ;P.)
[Readers who find the next section boring should skip below to "=== Substrate ===".]
I imagine I (sadly) thereby need to start by defending my history in the Android community, as many people seem to not be aware of much of it; it actually goes back very far, as I had promised the overall mobile community that if Android were ever rooted, I'd immediately start looking at it in earnest (before there was a device, I had already been messing around with the emulator, but the device concepts Google had at the time were more like slightly souped-up feature phones, not competitors to something like an iPhone). So, in 2008, when that first "root console attached to keyboard" mistake was found on the G1 that let you get a root telnetd running by just typing it into any search field, I dropped everything and drove two hours to Los Angeles to pick up a G1 (they were not selling them in Santa Barbara yet, due to T-Mobile not really having a presence here at the time). As promised, I immediately set to work attempting to help out.
As I ran a number of mailing lists already for iOS, I set one up for Android called g1-hackers, which attracted a good number of people, and even a few Google employees who worked on bionic and the kernel. On this list is where the G1's bootloader was first dumped: if you've ever heard the stories about Eddie Dost figuring out how to do it, this is that. In fact, it was from my G1, with a kernel I compiled (following Eddie's direction: I did not know much about flash drivers), that that first Android NAND was obtained (as Eddie had already updated his device and thereby didn't actually have root). Here is a link to the mailing list thread, directly to the post where we finally succeeded and I provided the kernel image I used so that others could perform the same dump on their own devices.
http://www.telesphoreo.org/pipermail/g1-hackers/2008-December/000096.html <- [g1-hackers] G1 boot code
Around that same time, I was also contributing to AOSP, providing a bunch of patches to things like mount and init, as I wanted to be able to get Android devices to a state where they could run something much closer to Debian than Android (I had my eyes set on kind of a hybrid). In the process of doing this, I wrote a guide that for a couple years subsequent were the canonical instructions for getting a bootstrapped build of Debian installed as a chroot under Android. At the time the patch turnaround on AOSP was sometimes over half a year (and almost never shorter than a couple months), which made contributing to the project sufficiently painful that I eventually stopped. If you search through Android's codebase, though, you still find some of my work.
http://www.saurik.com/id/10 <- Debian & Android Together on G1
At the time, I honestly do not remember XDA having yet become "the place" where people spent much time talking about Android: instead, a lot of conversation happened on IRC (which is where the iOS community had already been, and where it remains). There was a channel that I was a part of which included a bunch of people whose names would hopefully be familiar to people around here, including JesusFreke (and, much later, Cyanogen). I got to see the birth of a lot of great websites and tools (such as JesusFreke's smali/baksmali) while participating on that channel. Apparently, I was talking about "Substrate for Dalvik" on that channel in November of 2008 (which is also when I first joined XDA): that's how long I've been staring at this ;P.
During the next couple years, I ended up developing and maintaining a website called Cyrket, which had the mission to allow developers and users to search the contents of the Android Market using their desktop web browser. It also solved a few key problems that developers had with comments, in that you could only see comments for apps your device had access to that were then written in your language. Developers without devices, or with devices that could not see their product (which often included those that paid extra for the ADP1, which could not see copy-protected apps) could not see comments at all. Cyrket presented all of the comments for your application in all regions in all languages (and even used Google Translate to translate them all into your own language).
The way Cyrket had worked is that I scraped the contents of the Market using the same protocol Google's client used, indexed it (supporting find-as-you-type search), and exported it all to the site (well, originally, it was actually just a live client, but then it got really popular ;P). It got me into some mild trouble occasionally with the Android Market team, but overall no one seemed to mind it that much. Cyrket was actually the primary site people used for this purpose for a long time, and I even got the impression that people at Google were begrudgingly using it as it was more convenient than the alternatives. There were a few times where it had to be taken offline (due to changes and rate limits from Google), one time for months, but I'd usually figure out some new way to get it running. Honestly, though: I was really glad when Google finally launched a website for the Market and I was able to stop working on Cyrket ;P (and also glad that Google added most of Cyrket's features for developers to their publishing console, features that Apple actually still doesn't have AFAIK).
http://www.androidtapp.com/cyrket-android-market-browser-back-from-awol/ <- Cyrket Android Market Browser Back from AWOL!
Since those times, I mostly felt the need to get Substrate "awesome" (which started to really come together during 2011, after Cyrket was no longer needed), and so didn't do many larger projects on Android until recently. That said, I have been involved in things related to exploits and security. One of the higher impact things that I did was to release mempodroid, an implementation of the mempodipper exploit described by Jason A. Donenfeld for Linux 2.6.39+, which became the primary method to root devices running Android versions 4.0.0 through 4.0.2. Much more recently, users have been using Impactor, my implementation of the various "Master Key" exploits (based both on bugs described by Jeff Forristal as well as techniques I pioneered against a random AOSP bug).
https://github.com/saurik/mempodroid <- mempodroid README
http://www.saurik.com/id/17 <- Exploit (& Fix) Android "Master Key"
Given all of this, I hope people can get a feeling for just how strange and depressing it feels to me when people seem to suddenly believe I'm some kind of foreign invader . (FWIW, I also feel rather awkward having to describe all of this in this fashion, but frankly I'm at a point where I'm realizing that if I don't explain it in this much detail myself, no one else will. While I'm certain I'll get some people responding really negatively with comments like "he's such a blowhard, going on and on about silly little things he did", so far when I've given similar spiels to people in person at conferences, they often go "oh wow, I remember that tool/happening, but didn't remember that that was also you", and so figure that this might go a long way to fixing this weird problem: I'm not just "that iOS jailbreak guy".)
=== Substrate ===
Alright, now with that aside: in time for Google I/O (which was arguably bad timing, as I was then immediately unavailable for days ;P), I finally released Substrate. Substrate (in my clearly biased opinion ;P) is actually really cool: as far as I know, it is currently the only tool available for Android that allows developers to easily modify native code without patching/replacing. I know, for example, that people often ask how to modify features like the holo themes that are implemented in C, and the answer is Substrate: if you can find the code (which is often exposed via a symbol as there are tons of C++ symbols available on most Android builds) you can use Substrate to hook it at runtime in a way that avoids having to patch the files on disk, allows developers to deploy their changes across multiple ROMs, and supports the idea that users should be in charge of the specific features that they have on their devices (as opposed to ROM distributions).
As another concrete example that maybe makes this more obvious: sometimes you download a program from the Play Store (which, incidentally, I have a very hard time not constantly still calling the Android Market ;P) that is pretty much just a massive JNI binary--maybe an OpenGL game or a media player of some sort--that refuses to run on a device that has been rooted. A really common way that developers implement such checks is to do things like verify the existence of files on disk. The simple/common checks are very easy to detect and defeat using Substrate as you can hook the native "open" call from the C standard library, check if the filename is something like /system/xbin/su, and return "nope, not there".
http://www.cydiasubstrate.com/api/c/MSHookFunction/ <- MSHookFunction()
Substrate lets you do this kind of hooking in any system daemon (not just those spawed via app_process). Yes: if there's a program running in the background of your phone, some native service written by the OEM that manufactured the device, you can use Substrate to modify it. A lot of very interesting extensions on iOS involve these kinds of hack; for an extreme example, the software unlocks that we used to have for earlier iPhones involved modifying CommCenter, a native program that initializes the radio hardware: by hooking some of the code in that daemon, it was possible to, at just the right moment, inject a different command sequence over the serial connection to the baseband, exploiting it for the unlock.
http://www.cydiasubstrate.com/inject/android/ <- Android Native Injection
Of course, Substrate also supports hooking Java code (yes, a little like Xposed, which at some level uses the same underlying trick I walked people through in my talk at Android Open 2011). Somehow, though, a lot of developers don't seem to catch all that other stuff that Substrate lets you do, and get hung up on this one part that Xposed also manages, leading to all those aforementioned irritating comments about how "there's no point to Substrate because we already have Xposed": Xposed can't do most of the things Substrate can do (and the developer has even told me that he actively tries to avoid Substrate-like techniques as they are "pretty complicated", so it isn't even moving in that direction). FWIW, on iOS it took a lot of time for Substrate to get these features (it did not have them in 2008 when I first released it): they aren't trivial ;P.
http://www.cydiasubstrate.com/api/java/MS.hookMethod/ <- MS.hookMethod()
Even within the restricted context of modifications to Java, however, I think Substrate has a lot to offer. Again: I actively refused to release Substrate until I felt I had truly nailed a few things, including in particular the Java API (at Android Open 2011, I only supported JNI, which developers there told me would not lead to traction). I was a major proponent of aspect-oriented programming when I was younger, I got into byte-code engineering in college, and I co-published a paper on a Java code modification framework called jMonitor in 2004: this is something I've been thinking about for a long time, and I think the approach I take has some merit in and of itself. I know a lot more can be done (I feel it would be really interesting to have AspectJ-style pointcuts, for example, or the kind of bytecode-level instruction matching that I implemented as part of jMonitor <- features not described in the paper, I think ;P), but I felt a good first step was be to directly leverage the iOS community's six years of experience.
http://www.cydiasubstrate.com/id/6dfa187d-6e04-4f97-b63a-ae75b5338e01/ <- jMonitor [RV '04]
To this end, Substrate provides an API for Java that is very analogous to the API that it provides for modifying C/C++ and Objective-C. The focus is on "I know about some code and I want to modify it", allowing you to not have to think much about the timing or execution details of the program that may be loading that code (so you never have to think about "packages" or "processes" or "applications": you just concentrate on "classes", and thereby don't need a million "helper APIs" to handle each narrow timing case). To enable this, I use the aforementioned ability of Substrate to modify native code to hack features into the VM itself, giving me the ability to instrument events like "a class has been loaded". If you want to hook a method of a class from Apache Commons, and you want to hook that class no matter whether it was loaded as part of an application or dynamically as part of a classloader for a plugin downloaded by an application, this is trivial to express with Substrate. AFAIK, that use case isn't even describable using Xposed.
http://www.cydiasubstrate.com/api/java/MS.hookClassLoad/ <- MS.hookClassLoad()
This kind of VM-level modification and runtime code generation support (that is heavily flexed on iOS Substrate, and thereby has had years of in-the-field testing; so far Android has exposed just one bug in its ARM reassembler after release, and that was only in the qemu emulator for some reason) also means that Substrate's implementation of hooks is highly efficient: to compare again to Xposed, every time a method that has been hooked is called via Xposed, there is a linear-time search through a linked list doing a rather heavyweight comparison to determine which method it was after the fact; with Substrate, every call is direct, there are no lookups, and there are no comparisons, so you can hook an arbitrary number of methods with no slow down, so even very small methods that are called very often can be hooked without issue.
Additionally, with Substrate I wanted to address a specific pain point that many people would bring up when I'd give talks: "how is this secure, and how do I control what apps can use these features". This became even more important, as I wanted Substrate extensions on Android to be easily deployable via conventional means, such as the Play Store (yes: Cydia Substrate itself is in the Play Store, as I believe it is important for these kinds of features to not just be in the hands of developers on forums, but to be used by end users everywhere). To this end, I integrate into the Android security model, providing a special permission that applications must have to install a Substrate extension. This helps enable the idea that Substrate mostly "gets out of the way", becoming more of a technical detail behind your extension rather than something users will need to interact with constantly to activate or update your product.
I also wanted to provide at least something that would help solve the "reflection hell" that developers seem to always find themselves in while attempting to do runtime code modification in Java (even back on desktop Java using AspectJ). I thereby provide the means to "bless" a class loader, allowing it to access private fields and classes without the overhead of reflection: the access checks, for just that one class loader, no longer apply. Substrate extensions are loaded into such a "blessed" classloader. (I do not, even though I could, ever just whack an access check VM-wide; Xposed does this, and I feel like it is going to have security implications on Java security contexts applied to class loaders for plugins.) In the case of WinterBoard, for example, I don't ever have to deal with invoking Methods or getting Fields: setAccessible is just a dim memory.
Being able to use this functionality, however, can be awkward, and in some cases is almost impossible: while testing this feature, I realized that developers would end up needing "public stubs" for all the classes they were working with, but the calling convention for a public method and a private one is different, so the calls fail at runtime. I thereby ship as part of the Substrate SDK (yes, there's an easily-updated SDK package that you can download using the Android SDK Manager ;P) an extension to javac itself (as you might imagine at this point, written using AspectJ) that turns off access control checks: you can thereby access private fields or call private methods with no extra work both during development and at runtime. This all works sufficiently well that I generally run all of ant under the modification, such that anything ant compiles becomes "blessed".
http://www.cydiasubstrate.com/id/c17c554f-b603-4e3b-8f99-ebb3528e3ef8/ <- Java Access Controls
(And yes: this is one of the things that caused Substrate to get delayed even longer than it already had been. There was also a rather serious delay caused by my attempts to really nail the boundary between "code that is shipped with Substrate" and "code that is shipped with the extension", something that burned me a lot throughout 2013 as it was the kind of problem that spending time actively thinking about didn't directly help, requiring an epiphany I had soon before Google I/O. Arguably had I been willing to ship without documentation at all, and had I generally cared less, I would probably have had everything out in very early 2012, but during January-May I started working on the initial draft of cydiasubstrate.com, as I had apparently incorrectly thought that such efforts would be critical to developer adoption.)
Again, I write this in the hope that it clears up misconceptions, either about myself or about Substrate. As far as I can tell, Substrate has a lot of very unique value propositions: things that currently are only made possible by Substrate; and, even within the restricted scope of hooking Java code inside of a service being managed by Zygote (the only area of overlap with Xposed), I think that it offers a bunch of advantages in security, performance, deployment, and ease of development that cannot be so casually dismissed with a flippant "we already have Xposed (go home)". A lot of these features (and I haven't even gone into all of them: I could write paragraphs about the advantages of how Substrate's API handles chained hooks, the ways I enable extensions that need to cross classloader boundaries, or the way Substrate makes it easy for end users to temporarily disable extensions without complex tooling) come from having spent over a decade now thinking about this problem and the last five years actively managing a developer ecosystem with tens of millions of users on iOS.
I am thereby happy to answer any questions about how to use Substrate, issues with Substrate on any device (I never blame the device: I might not have a fix immediately for a specific problem, but I always consider it Substrate's job to work around issues the device throws at it to get its functionality in place so the task will at least end up on my todo list), or even about me (as a lot of why I find writing this both so important and so painful are due to the occasional-yet-present more-personal attacks/misconceptions I often seem to receive about somehow being an "outsider"). (That said, please do have some patience: sometimes my ravenous need to do nearly 24/7 testing on a specific device has to give way so I can go to a conference I'm giving a talk at, or so I can focus on a different problem that might be more pressing or simply have a higher probability of near-term success: spending an infinite amount of time on one problem is unfair to all of the other problems that exist ;P.) [And, in fact, I have a meeting I have to be at tonight, but which hopefully won't take insanely long.]
Reserved Post
["reserved", as apparently you always should have at least one of these ;P]
Links to Extension Threads
[and finally, I can see ending up with a page that might link to other threads on XDA, although arguably I should put this on cydiasubstrate.com. right now, most projects that use Substrate are in Play. I am not certain if I'm now just misunderstanding how to use XDA, though: again, this is my first thread I've started myself]
Wow. The timing couldn't be any more perfect for you to post this.
I do not have an Android device yet and have been theorizing exactly how I could easily make modifications to applications.
Because I am just getting started in the Android development community, I don't have any biases towards one framework or the other.
Sooo.... this is on my watch list.
gugbot said:
Wow. The timing couldn't be any more perfect for you to post this.
Click to expand...
Click to collapse
The opinion of many (reasonable) people differ ;P.
gugbot said:
Sooo.... this is on my watch list.
Click to expand...
Click to collapse
Yay! If you have a moment, I'm curious: how/why did you find this thread? It seems like very few people actually go to this "Frameworks" sub-forum; there are almost no threads posted to it except the one about Xposed, which I'm presuming people must be finding by links from other places (whether random websites or other threads on XDA).
saurik said:
The opinion of many (reasonable) people differ ;P.
Yay! If you have a moment, I'm curious: how/why did you find this thread? It seems like very few people actually go to this "Frameworks" sub-forum; there are almost no threads posted to it except the one about Xposed, which I'm presuming people must be finding by links from other places (whether random websites or other threads on XDA).
Click to expand...
Click to collapse
I was browsing in development tools and was surprised to see that a Saurik posted about Cydia Substrate!
I was brought to this forum by one about theme development?... Maybe you should post this in a forum with more traffic. There seems to be an endless amount of categories for everything.
i have try your cydia substrate on cm10.1.3 stable..device samsung i9300..
install winterboard..apply icon pack but icon pack not applied..
then when want to open other apps the apps fc..except winterboard..
slipar said:
i have try your cydia substrate on cm10.1.3 stable..device samsung i9300..
install winterboard..apply icon pack but icon pack not applied..
then when want to open other apps the apps fc..except winterboard..
Click to expand...
Click to collapse
Yeah, as I mention in this thread WinterBoard was more of a demo that has been difficult to justify improvements to . This isn't an issue with Substrate, at least.
Would you mind sending me the crash report from the adb log? At least, would you mind telling me the name of the theme you applied? Also, thinking about it, CyanogenMod already has a theme engine... it never occurred to me how WinterBoard would interact with the existing theme engine in CyanogenMod (although I guess thinking even longer about it, I see no reason why it would fail horribly... it should just layer on top).
saurik said:
Yeah, as I mention in this thread WinterBoard was more of a demo that has been difficult to justify improvements to . This isn't an issue with Substrate, at least.
Would you mind sending me the crash report from the adb log? At least, would you mind telling me the name of the theme you applied? Also, thinking about it, CyanogenMod already has a theme engine... it never occurred to me how WinterBoard would interact with the existing theme engine in CyanogenMod (although I guess thinking even longer about it, I see no reason why it would fail horribly... it should just layer on top).
Click to expand...
Click to collapse
hope i send u the correct logcat..
im using ios7 concept theme..g play link here
slipar said:
hope i send u the correct logcat..
im using ios7 concept theme..g play link here
Click to expand...
Click to collapse
Thank you so much for the information. Here is a new version of WinterBoard that seems to work with this theme.
http://cache.saurik.com/apks/com.saurik.winterboard_0.9.3922.apk
thanx saurik..tested but this time winterboard just fc when try to change theme..
logcat attach..
slipar said:
thanx saurik..tested but this time winterboard just fc when try to change theme..
logcat attach..
Click to expand...
Click to collapse
I'm sorry about that issue... this is actually quite interesting to me as it might indicate that I need to do some more work on the blessed compiler as it relates to miranda methods. I had verified that the theme functioned, but had not gone back to attempt to re-verify the setup activity itself, which I guess hadn't been recompiled in a long time. I've added a temporary workaround to the issue while I investigate further. ("Humorously", if you have Xposed installed, I am pretty certain that the WinterBoard settings activity would have worked, as Xposed just destroys the access control checks for the entire VM.)
http://test.saurik.com/xda/com.saurik.winterboard_0.9.3922+1.gf733f01.apk
Hey there, I just happened upon this thread while deeply perusing the boards after just getting home from a 17hr drive and being unable to go to sleep yet. I am VERY interested in the substrates capabilities, it sounds like a very interesting concept. I am a new developer and am wanting to learn more and play more....I use xposed on my phone now and was considering starting to develop modules for it, buuuttt I think I just changed my mind I'm on an att sgs4 running a 4.3ge Rom. Going to install the substrate the night via Play Store and mess around with it starting tomorrow. Thanks for this
Sent from my GT-I9505G using Tapatalk
Sc4ryB3ar said:
I'm on an att sgs4 running a 4.3ge Rom. Going to install the substrate the night via Play Store and mess around with it starting tomorrow. Thanks for this
Click to expand...
Click to collapse
Yay! (Now, watch your GT-I9505G be one of those few Samsung devices Substrate detects as incompatible ;P. Samsung has so many model numbers that all map to the same high-level marketing names that it's difficult to keep track of what's what. If that happens, and you are interested in helping out, I can implement one of my alternative injectors quickly for you to work with.)
saurik said:
Yay! (Now, watch your GT-I9505G be one of those few Samsung devices Substrate detects as incompatible ;P. Samsung has so many model numbers that all map to the same high-level marketing names that it's difficult to keep track of what's what. If that happens, and you are interested in helping out, I can implement one of my alternative injectors quickly for you to work with.)
Click to expand...
Click to collapse
It installed just fine, quickly and with no apparent issues
winterboard, however rendered neither theme I chose correctly, wondering if its the themes though.... Didn't get a logcat and then I hosed my system last night messing around too much, so I started fresh and haven't gotten back to substrate and wb yet....I'll be back to it withing a couple of hours
Sent from my GT-I9505G using Tapatalk
substrate source code
Saurik,
I've been dabbling some with Cydia Substrate and it seems to offer a lot of unique possibilities for Android apps.
Do you have any plans to release the source code for this like you did on iOS? I'd be very curious to learn more about how it works. Also, is there a link to your talk from the Android Open conference?
Thanks,
Fred
(Ugh. I have no clue how people keep up with a forum, especially with the website as slow to load every page as it is ;P.)
fjones8856 said:
Do you have any plans to release the source code for this like you did on iOS? I'd be very curious to learn more about how it works.
Click to expand...
Click to collapse
I currently do have an intention to release the source code, but I'm not certain under what license (all of the licenses I normally use don't solve the specific issues related to Substrate). That said, no one seems to care much about Substrate on Android: on iOS people tend to (almost to a level of it being a problem) jump on new solutions to evaluate constantly, whereas on Android people seem to just snark "we already have X" even when there are compelling advantages to a replacement. Given this situation, I am highly unmotivated to spend the time to figure out the right solution, given that in a way Substrate is "my magnum opus": it is the culmination of the research and experience of so many years of my life, that passing up the ability to license it to the companies that sometimes talk to me about that (for either enterprise wrapping or security) to satisfy a group of people who are mostly asking for the source code specifically to replicate the technique *and then avoid using Substrate*, makes very little sense.
On the project side of it, Substrate on iOS only ever received a single code contribution from someone I wasn't already so close with that I was sharing code already. It isn't even the kind of project that one would expect getting many contributions: it is more of a backend technology, and the extent to which it has a GUI is actually a bug (I intend for it to be 100% seamless as part apps that use it: Substrate on iOS does not have a GUI and never will have a GUI, and that's how I think it really should work on Android as well, but of course right now I need the silly Install button). If anything, on iOS, we often end up with random companies that want to "own the scene", which ends up with them forking Substrate in ways that cause platform incompatibilities for other developers: Substrate on iOS has thereby actually been closed source now for almost two years, and it has actually improved the stability of the platform. I thereby am somewhat loath to "repeat the same mistakes from before" and end up with forks.
fjones8856 said:
Also, is there a link to your talk from the Android Open conference?
Click to expand...
Click to collapse
There was no recording of the actual talk, just of the keynote introduction that I already link to from my website. In the talk I walked people through a demonstration of using an early version of the JNI-level Substrate API, and showed how it worked (which was very simple at the time). In essence, I demonstrated, with my exact code on the projection, the technique that Xposed started using half a year later (which is just "oh, I'll change the contents of this Method object, as apparently the runtime doesn't care if the Method is allocated as part of a Class; if I do it right I can simulate registerNatives") and the most obvious way of implementing MSJavaHookClassLoad (which--for the really really low-level API I had at the time, on pre-4.0 VMs that didn't have complex JNI stacks--is clearly "MSHookFunction the class load and provide a callback"). Everything is going to be new for ART, though: the techniques are going to have to be much more sophisticated (which I'm excited by, as this is a game changer).
Pm sent
Sent from my GT-I9505G using Tapatalk
A few weeks ago, I posted a very unfortunate Google+ post of the creator of Focal and why it was removed from the CM codebase. It was a depressing story and it really started to make you wonder about where CM is going.
This time, after reading an extremely well-written article, I've come to a similarly depressing conclusion: Android by Google is slowly becoming as locked down as iOS, but not in the sense that you think; it's not about what apps let you do what, it's the developers.
We've finally arrived at a critical flaw with the way Android is developed and these days, I can no longer claim that Android (by Google) is "open" anymore.
Feel free to give this a read (Disclaimer: I am not affiliated with Ars Technica in any way).
http://arstechnica.com/gadgets/2013...ntrolling-open-source-by-any-means-necessary/
It's not just about Amazon's version of Android; CyanogenMod is for all intents and purposes a "fork" of Android. It is designed to work without Google Apps and as we all know, we flash those seperately. But that's the problem, the answer isn't just "Well, I'll just flash the Gapps and it will work like it should". What will happen if new Play Store apps start referring to features in the framework that don't exist in a form that we can flash? What if the license to flash the Gapps gets revoked?
How will CyanogenMod start adding features to apps that were originally AOSP but are now closed source? What will happen when the open source Messaging app is abandoned and turns into a Hangouts feature? How can CM stay on top of that?
It's not as simple as "take the source we currently have and work with it", because what will happen when Google adds a killer feature to an app that depends on some API that is no longer open source?
These are some rather frightening questions to deal with. I don't know where Android is going, but I'm certainly starting to wonder what's going to happen to it.
I'd appreciate any and all input on this.
Not very continuous, but here's my thoughts about the article:
The Gapps license is meant to lock the makers of Android phones into Google, so users get locked within Google and Google can gain revenue from the users. After going to that extent to make sure Google gets to keep the device's user, what's to gain if Google users of the device who flash CM to be locked out of the system instead of keeping them "trapped" with the Google ecosystem even with a non Google ROM? Doesn't make any sense does it?
I suppose we will still have to flash them like we flash the Play Store now. Unlike Amazon, CM (for now) actually still relies on Google and doesn't "divert" revenue to another company and therefore Google would be more than happy to let their apps be used. But if CM does start going the Amazon way, I believe Google may lock CM out.
Those APIs take time to develop, take the Maps API for example - you think they spent millions, if not billions mapping the entire world and even roaming every street just to make sure you can find your way around for free? They'll need to recoup their costs somehow.
While Android is open source and contributed by Google for free, don't forget Google is a company, not a charity. They have to make money or their shareholders won't be happy. Even if their shareholders are massive fans of open source they also have thousands of employees to pay, and all that costs money. And don't forget, when a company is providing free stuff for you to use, you are not their customer - you are their product. Android will change in ways that will keep Google profitable and keep competitiors unprofitable, while keeping the users as comfortable as possible so they will continue to be their product.
cccy said:
Not very continuous, but here's my thoughts about the article:
The Gapps license is meant to lock the makers of Android phones into Google, so users get locked within Google and Google can gain revenue from the users. After going to that extent to make sure Google gets to keep the device's user, what's to gain if Google users of the device who flash CM to be locked out of the system instead of keeping them "trapped" with the Google ecosystem even with a non Google ROM? Doesn't make any sense does it?
I suppose we will still have to flash them like we flash the Play Store now. Unlike Amazon, CM (for now) actually still relies on Google and doesn't "divert" revenue to another company and therefore Google would be more than happy to let their apps be used. But if CM does start going the Amazon way, I believe Google may lock CM out.
Those APIs take time to develop, take the Maps API for example - you think they spent millions, if not billions mapping the entire world and even roaming every street just to make sure you can find your way around for free? They'll need to recoup their costs somehow.
While Android is open source and contributed by Google for free, don't forget Google is a company, not a charity. They have to make money or their shareholders won't be happy. Even if their shareholders are massive fans of open source they also have thousands of employees to pay, and all that costs money. And don't forget, when a company is providing free stuff for you to use, you are not their customer - you are their product. Android will change in ways that will keep Google profitable and keep competitiors unprofitable, while keeping the users as comfortable as possible so they will continue to be their product.
Click to expand...
Click to collapse
First, I appreciate the input! I was looking forward to intelligent discussion and it's great that the first reply is just that.
I would like to clarify though; my concern is not so much about Google making money; they are a business and deserve to make money in whatever way they see fit. We have something they want (ad clicks and search history) and as long as they provide an experience worth using, I don't mind that transaction at all.
My worries start with what the custom development scene will look like one or two years from now if the base apps that make Android useful on its own (and by extension, useful to custom developers) have been molded into Google Play apps or frameworks or APIs.
In parallel, it's also starting to make sense why Cyanogen continues to put effort into alternate applications such as Apollo and Focal; they saw this coming way before we did.
LiquidSolstice said:
First, I appreciate the input! I was looking forward to intelligent discussion and it's great that the first reply is just that.
I would like to clarify though; my concern is not so much about Google making money; they are a business and deserve to make money in whatever way they see fit. We have something they want (ad clicks and search history) and as long as they provide an experience worth using, I don't mind that transaction at all.
My worries start with what the custom development scene will look like one or two years from now if the base apps that make Android useful on its own (and by extension, useful to custom developers) have been molded into Google Play apps or frameworks or APIs.
In parallel, it's also starting to make sense why Cyanogen continues to put effort into alternate applications such as Apollo and Focal; they saw this coming way before we did.
Click to expand...
Click to collapse
I believe the custom development scene wouldn't get affected much. After all, remember the old XDA-Developers? Windows was all locked down, but the cooks still managed to make customized ROMs. What's more, Google wouldn't want to lose their "products" - Google wants us to continue to use their services so they can earn money, they wouldn't lock us out.
What competitors lack is the capability to access Google's services (Frameworks, APIs, etc) as Google has ways to block them (Which is why we had circumvents like device spoofing). If you had a device designed for Google's version of Android, I am sure Google would still enable access if you use a custom ROM. The point of locking those competitors out is to force them to embrace Google's version of Android and not use their own forks which would keep Google out of certain aspects of the user's phone, decreasing revenue. Therefore, if you could roll your own custom ROM, it makes sense for Google to continue supporting you so you still completely rely on them instead of "outsourcing" to other competitors.
CM puts effort into alternate applications because as you can see right now, CM's starting to roll their own commercial forked devices - what happens after that? If you have seen the ways of other commercial versions of Android (Amazon, China brands, etc), they start replacing certain revenue generating aspects of the phone to use their own service instead of Google's. Certainly not what Google wants.
In short, I would say, if you are a small custom ROM user, Google isn't going to come after you, they want you to use their services! But if you are a competing company, expect your devices to be locked out from Google in the hopes that they eventually force you to bow to them and convert all your users completely to Google's "products".
This may be stupid, but I couldn't find any resources regarding this. We have custom recoveries for android devices but why isn't there custom bootloaders like there is for PCs ? Like in the PC space we have the likes of reFind and gnu grub.
Thanks
There are some instances of alternate bootloader projects. Just that they are not popular,
[Bootloader] LK for Xperia T
LK for Xperia T LT30p Only - Unlocked Bootloader Required WARNING 1: This modification makes changes to the devices partition table. I (lilstevie) am not responsible for any damage to your device or data loss that may occur. WARNING 2: ICS...
forum.xda-developers.com
EFIDroid
EFIDroid is a easy to use, powerful 2ndstage-bootloader based on EDKII(UEFI). It can be installed one-click with the EFIDroidManager app. You can add/remove/edit multiboot ROM's. There's no special support needed by ROM's or RecoveryTools(no...
forum.xda-developers.com
The developer of EFIdroid stopped developing in 2019.
efidroid on Android 9 and 10 devices ? · Issue #152 · efidroid/projectmanagement
Hi, I just want to know if efidroid supports devices with 6 GB RAM and 64/128 GB Storage devices running Android 9 and Android 10 ? thanks.
github.com
Not to mention you would need OEM's to cooperate....
Thanks @karandpr for that github comment a lot of info there. Thanks @galaxys too. So a quick summary would be that the reason is that for the bootloader to work smoothly there has to be support from the kernel too, which the OEMs should do and probably would not. But I didn't think about the support in the kernel was an issue. That does seem to be a lot of work and I see the reason now.
al_l_en said:
Thanks @karandpr for that github comment a lot of info there. Thanks @galaxys too. So a quick summary would be that the reason is that for the bootloader to work smoothly there has to be support from the kernel too, which the OEMs should do and probably would not. But I didn't think about the support in the kernel was an issue. That does seem to be a lot of work and I see the reason now.
Click to expand...
Click to collapse
I don't think Google intends to open up android anymore. They want restrictions like iOS but pretend to be open source for the "goodwill". What's the use of AOSP if you cant effectively install it on a device or your important apps don't work?
I believe PinePhones are the ones that can have truly open-source compatible hardware. The specs are underwhelming but the community is really good.
You can get spares easily and the battery is removable.
Only thing is they are mostly out of stock.
karandpr said:
I don't think Google intends to open up android anymore. They want restrictions like iOS but pretend to be open source for the "goodwill". What's the use of AOSP if you cant effectively install it on a device or your important apps don't work?
I believe PinePhones are the ones that can have truly open-source compatible hardware. The specs are underwhelming but the community is really good.
You can get spares easily and the battery is removable.
Only thing is they are mostly out of stock.
Click to expand...
Click to collapse
Yeah those are great but the problem is that they are not usable for "normies" which will prevent mass adoption and hence cannot have a sustainable business model.
But I think google is not the only one to blame, like couldn't the OEMs actually provide bootloaders that can boot signed os images. Or is there any technical or security difficuties in doing that.
al_l_en said:
Yeah those are great but the problem is that they are not usable for "normies" which will prevent mass adoption and hence cannot have a sustainable business model.
But I think google is not the only one to blame, like couldn't the OEMs actually provide bootloaders that can boot signed os images. Or is there any technical or security difficuties in doing that.
Click to expand...
Click to collapse
Normies are afraid to change the default browser, so bootloader is really out of their leagues.
Phone tinkering is a hobby, not a necessity. Phone tinkering itself is not a sustainable model.
Google is to blame primarily. Because they have a stringent list of requirements for devices to pass CTS. You can read the bootloader requirement and judge yourself.
Android 11 Compatibility Definition | Android Open Source Project
source.android.com
Without passing CTS, devices cannot use Google apps, they cannot get push notifications and they cannot pass SafetyNet checks used by most banking apps.
At the end of the day do I want to spend 100s of hours to bring a feature to an android phone which will probably be used by 10 users and deprecated by the time I finish doing it?
or do I want to buy a phone which will allow me to tinker freely in a community and ecosystem which allows modification?
For our tinkering pleasures, Pinephone is the way to go for now. They have support from Manjaro, Debian and KDE. Which is a big thing IMO.
Or else there you can roll your thing in RaspberryPi?
While going through related details I found an article about google probably switching to hardware based safetynet checks which could be ending google play compatibility on custom roms.
It really seems like google is using security as an excuse to make sure that there are no competitors in their business space.
Maybe this is because I have been only doing web development and only started learning app dev, but the reasons google use for CTS like for enforcing DRM, is also handled on websites while allowing openness and being neutral (or maybe the web is not as secure as something like this, so forgive me if I am wrong). Android could really take pages off the web ecosystem for being a neutral platform.
I really appreciate the patience for hearing out and also the references(and the rabbit holes that it was followed by) really taught me a lot about general android architecture.
al_l_en said:
While going through related details I found an article about google probably switching to hardware based safetynet checks which could be ending google play compatibility on custom roms.
It really seems like google is using security as an excuse to make sure that there are no competitors in their business space.
Maybe this is because I have been only doing web development and only started learning app dev, but the reasons google use for CTS like for enforcing DRM, is also handled on websites while allowing openness and being neutral (or maybe the web is not as secure as something like this, so forgive me if I am wrong). Android could really take pages off the web ecosystem for being a neutral platform.
I really appreciate the patience for hearing out and also the references(and the rabbit holes that it was followed by) really taught me a lot about general android architecture.
Click to expand...
Click to collapse
Theoretically, Google can end GPlay compatibility on Custom ROMs anytime they wish. It's just that lot of App Developers don't use SafetyNet the way it is intended and Google doesn't roll out its strict check. They do it once in a while.
They don't have any competitors in their business space. It's a very well-thought monopoly.
CTS restricts Google Play API access to vendor operating systems. So vendors like Samsung, OnePlus and others have to play by their rules. IIRC, the cost of Play API is around 15$ per device but it is subsidized for large quantities.
End users don't really care about Play API. But App Developers do.
Without Play services, there is no easy way to integrate push notifications, ads, maps, analytics, metrics, and so on. Rolling your own thing will take years to develop and won't work as seamlessly as the play service counterparts.
I don't think Google will ever cede their monetary interests for open collaboration.
karandpr said:
I don't think Google will ever cede their monetary interests for open collaboration.
Click to expand...
Click to collapse
Yeah that's for sure. The only way this monopoly can break is when an opensource alternative to google play services and other apis exist and while doing that it must be compatible with the existing google apis. And that is probably not going to happen in a long time. Although microg does solve this to some extent, but still it is a second citizen.
Some of the functionality is already there, like most of the google apps like docs and drive could replaced by nextcloud and then maps could be replaced by osmand. If some company, preferably an OEM, comes and integrates all of these into a package maybe there's hope. I think /e/ os tries to do this to some extent.
You might find this resource useful. As they have gone over a comprehensive set of bootloader software and tried to outline their primary features in detail. Hopefully, you’ll be able to determine the best one for your use case. https://www.ubuntupit.com/best-linux-bootloader-for-home-and-embedded-systems/