Live data from Hacker News

Keep Android Open

keepandroidopen.org

881–890 of 907 posts

Re: Keep Android Open

#881

Earlier quoted context omitted.

Having a rooted android 11 phone for years was never a problem. My bank apps worked just fine. Even for work stuff (usually). It's on the personal side where I actually started to value having a virtual credit card on my phone with Google pay or apple pay. The stack to enable that securely is only on android and iOS and there's nothing else out there that has that. Open source community needs a full stack for attesti…

Seconded. The NFC payment feature is useful on mobile in a way that generic "online banking" just isn't IMO. In the same category are transit apps, ride-hailing apps, social messaging, and a (very) few other others. The problem is that payment really does require a secure stack, as you describe.

I prefer to use an actual credit card, in order to keep the control over my computing in my hands.

Re: Keep Android Open

#882
post #749
post #665

Earlier quoted context omitted.

Last I checked the situation was similar to what it is in Calyx, which is that it's not officially supported and you have to keep manually reapplying the root after every update.

Userdebug builds of GrapheneOS with ADB root access are officially supported. We recommend setting ro.adb.secure=1 rather than making a standard userdebug build with always-on unauthenticated ADB if it's not solely for development. Modifying the official builds by replacing part of the core OS with Magisk and then using that to modify the rest of the OS dynamically is what's not officially supported and strongly disc…

I've been trying to make a UserDebug build for the last few days by following the official instructions at https://grapheneos.org/build, and am running into some trouble which I suspect are due to minor steps in the documentation that are missing or incorrect.

Is it possible to get some help on this? Posted some messages to the #general and #development channels on Discord (you mentioned in your very helpful comment at https://news.ycombinator.com/item?id=42536302 that's the go-to for support) but am not getting any response. I'd even be happy to make a fat donation or pay for some support here.

Also out of curiosity, have you guys considered providing an official release that is configured this way, perhaps gated behind a giant "Proceed at your own risk" banner? Is your reasoning here influenced by a perception that providing root access without enough friction might weaken your bid for Play Integrity recognition or your chances of individual app developers like banks treating your OS on equally trusted footing as Google's?

Re: Keep Android Open

#883
post #58

Back in the 2007 or when it came out in Sweden I bought the iPhone and started developing for it. This was cool, new and exciting and it was fine as long as my company was paying the $100 fee every year. But then I switched jobs and worked at a company which produced mostly open source code. Suddenly I would have to pay $100 every year just to be able to put my own software on the phone ... This is why I switched to…

You don't need a paid dev account to build and run on your own iPhone. I didn't have one most of the time

Re: Keep Android Open

#884
post #145

While I understand the reasons behind this campaign, I have mixed feelings about it. As an iPhone user, I find it frustrating that deploying my own app on my own device requires either reinstalling it every 7 days or paying $100 annually. Android doesn't have this limitation, which makes it simpler and more convenient for personal use. However, when it comes to publishing apps to the store, I take a different view. I…

Litmus test: Can you get NewPipe or other Youtube clients onto an Android phone? This is non-malicious software that users want to run but could reduce YouTube's profits.

Re: Keep Android Open

#885
post #46

Earlier quoted context omitted.

Most vendors (at some level) allow flashing custom distributions, as long as you didn't buy that device from carrier: https://github.com/zenfyrdev/bootloader-unlock-wall-of-shame... You will lose DRM-based apps (e.g. Netflix), Payment apps, and bank apps though.

Even phones from Motorola require you to literally ask permission to unlock your bootloader via a form on their website, which they then unlock remotely or you enter some generated code. Other manufacturers do the same, where you have to wait a period of like 45 days before being able to unlock, and then have to ask permission on their website to unlock your bootloader.

iiuc that is because malicious actors were buying phones in bulk, flashing them with backdoored/malicious operating systems, then re-selling them to people.

Re: Keep Android Open

#886

Earlier quoted context omitted.

To be fair, for "anything over 5 years old" you can probably find a privilege escalation exploit.

That might get you root but not a bootloader unlock.

Many of them are actually not bootloader locked.

iiuc the OG Verizon Pixel has an unlockable bootloader, but the operating system doesn't let you unlock it, meaning root access should allow unlocking the bootloader.

some devices have a legitimately locked bootloader, which means you're SOL.

Re: Keep Android Open

#887
post #746

Earlier quoted context omitted.

> There are no malicious apps in GNU/Linux repositories. That's definitely not the case. There have been repeated cases of developers shipping malicious code which ended up in distribution package repositories. Defining malicious is difficult and incredibly privacy invasive behavior is often not considered to be malicious. That software is also generally being used without a mandatory app sandbox with a proper permis…

> incredibly privacy invasive behavior is often not considered to be malicious This is not true for Debian, which is the upstream of PureOS. > therefore not actually significantly reducing trust in the upstream projects And yet, it has practically negligible number of malicious apps, especially compared with Google Play. It's far from perfect, and you are right that the sandboxing should be further improved. Neverthe…

> This is not true for Debian, which is the upstream of PureOS.

Lots of the software they provide has privacy invasive behavior and far more than that has poor privacy.

> And yet, it has practically negligible number of malicious apps, especially compared with Google Play.

Google Play is not the only app repository for Android-based operating systems. There are repositories in the style of traditional Linux distributions and also better approaches available.

> Nevertheless, it is a security model working in practice for a large userbase of Debian.

No, it has very poor privacy and security.

> It works especially well for technical users.

Being technical doesn't address the massive privacy and security issues. It only makes it less likely people install blatant malware instead of it being a problem through supply chain attacks and very poor security throughout the OS.

Re: Keep Android Open

#888
post #739

Earlier quoted context omitted.

Eylenburg's site has comparisons between a bunch of different types of software and services with a significant focus on privacy and security rather than aesthetic customization features, etc. For the Android comparison, GrapheneOS is the only privacy and security hardened OS included in the comparison. DivestOS used to be included before it was discontinued. An OS not including Google Mobile Services and branding it…

> For the Android comparison, GrapheneOS is the only privacy and security hardened OS included in the comparison. DivestOS used to be included before it was discontinued. An OS not including Google Mobile Services and branding itself as private based on that is a much different thing than a privacy and security hardened OS. Which other Android-based hardened OS could be included in the comparison? I was arguing on th…

> I was arguing on the other axis. It's got good coverage of OS options, but the list of features is indistinguishable from someone saying "okay, this is what GOS does; how do others compare to each of its selling points?"

That's not what they're doing. Their focus is privacy and security, but they've only included a single OS in the comparison aligned with that focus. They don't want to include less known operating systems such as forks of GrapheneOS.

> There is a difference between not shipping something by default, and being actively hostile to it.

GrapheneOS doesn't do anything that's hostile towards users having root access. It provides official support for it in userdebug builds, which we recommend doing with ro.adb.secure set to 1 rather than using the default Android userdebug configuration.

> Agreed, that would be foolish. Thankfully, nobody is suggesting that. Just use a permission prompt, like every android root solution has for... over a decade? I don't think I've ever seen anyone not putting root behind a permission prompt, actually.

That's not addressing what was said. Having a permission prompt to grant unconstrained root access to apps still involves giving root access to a massive portion of the OS, greatly harming the security model and losing a bunch of the security features. It inherently regresses security in an enormous way which having user-accessible root access alone does not. Making it accessible to apps through the high level OS UI is a far different thing from having a very well protected way for users to have root access.

> On, say, my laptop, I can configure arbitrary VPNs and firewall rules, and I can configure them independently.

There are major issues with VPN leaks without these things tightly tied together and having multiple things configuring firewall rules is going to cause major conflicts. Android provides support for having a per-profile VPN with built-in leak blocking where the OS takes on most of the responsibility for it and can properly integrate it everywhere that's required.

> Android conflates them such that - if not using root to work around the official way - your firewall app and your VPN app must be the same app.

No, it does not have to be the same app. It's a choice the app developers are making to not provide an extensible system interoperable with other apps but rather to do everything themselves. It's also not a conflation of these things but rather it needs to be implemented as an OS API to do it properly.

> It's nice that RethinkDNS has specifically added wireguard support to its firewall app, but the fact that they needed to is a symptom of a poor design.

It's not a poor design but rather a much better design which avoids the very broken approach of multiple applications including VPNs trying to configure firewall rules. It's impossible to get right without all of those applications closely coordinating which is not what happens. Having VPN support built into the OS with a well defined API for apps to implement it with leak blocking support built into the OS is a far better approach which can be done correctly instead of the mess on laptops/desktops you're talking about. Android doesn't have a great implementation of all this upstream but the high level design approach is a far better one. It needs a major overhaul to heavily use network namespaces instead of the current messy approach built on a legacy design. That does not change that providing the VPN leak blocking within the OS and an API for VPN implementations is definitely a better approach. Built-in WireGuard support would be nice too but should be done properly.

Giving root access to a huge portion of the OS including the high level application layer and to actual apps running on top is always an insecure approach to this. It's a lazy shortcut to skip having to implement a proper system following the principle of least privilege where the GUI portion of the OS does not simply have unconstrained root access to manage everything at a low level in incorrect, racy ways. The correct place for firewall management is netd with user-facing interfaces which end up controlling what's done by netd. netd has CAP_NET_ADMIN and is in fact contained by SELinux without access to much more than networking configuration. Giving unconstrained root access to the GUI portion of the OS where minor UI bugs can give permanent root access to an attacker where there's no way to revoke it if they want to stop it is definitely not the right approach to implementing more user-facing firewall controls.

Re: Keep Android Open

#889
post #775
post #738

Earlier quoted context omitted.

Eylenburg's site is focused on privacy and security for the comparisons. GrapheneOS is the only privacy and security hardened OS included in the Android-based OS comparison. None of the other operating systems listed in that comparison keep up with Android privacy/security patches or provide significant OS level privacy or security improvements. Many GrapheneOS features aren't listed by the table or are grouped in hu…

So, what you are saying is that Lineage has bad security because they are doing their best to support old devices as long as possible? Interesting position. It is a valid criticism but brings its own problems.

No, that's not at all what was said.

Pixel 7 is a fully supported device with Android 16 QPR1 available for it via the stock OS. /e/ is on Android 13 on the Pixel 7 without the kernel, driver and firmware updates released for it from the Android 14 launch and later. /e/ is a fork of LineageOS with drastically worse privacy and security.

LineageOS rolls back the security model and features a fair bit but does it far less than /e. LineageOS doesn't implement major privacy and security enhancements so a comparison of non-standard improvements in those areas is going to show that. LineageOS lags behind on providing the AOSP and vendor security patches, but far less than /e/.

The table covers the patch delays for both AOSP patches and device patches. It's showing how far behind they are on kernel, driver and firmware patches released for a device, not anything to do with end-of-life devices. Being multiple years behind on those patches for the Pixel 7 on /e/ has nothing to do with supporting an old device as long as possible.

Re: Keep Android Open

#890

Earlier quoted context omitted.

Seconded. The NFC payment feature is useful on mobile in a way that generic "online banking" just isn't IMO. In the same category are transit apps, ride-hailing apps, social messaging, and a (very) few other others. The problem is that payment really does require a secure stack, as you describe.

I prefer to use an actual credit card, in order to keep the control over my computing in my hands.

Indeed, I do too. But since you always need at least one backup means of payment, I keep a second virtual card on mobile for that. Which alas is a very convenient solution.
Post reply on HN