> One more time, what mobile phone owners need is an intermediary layer in the mobile OS where they can either manually or programmatically (via e.g. a random number generator) lie to higher-level application layers about the phone's data.

On Android, if you install Xposed, there's XPrivacy [0], which essentially had this premise. Unfortunately, it only supports versions 4.0.3-5.1.1 (and is no longer maintained).

I haven't been following the rooting community for a while, so someone can probably correct me if there's a better alternative now, but I believe the replacement is XPrivacyLua [1].

[0] https://github.com/M66B/XPrivacy

[1] https://www.xda-developers.com/xprivacylua-xposed-module-pri...

You can use XPrivacyLua on later Androids: https://forum.xda-developers.com/xposed/modules/xprivacylua6...

Many applications force-stop when you use Xprivacy. Even worse, applications like instagram refuses to login on a phone with xprivacy installed.

The fact that this is possible is a symptom of the wider problem - the OS should ensure that it's impossible for the other app to know if Xprivacy is running or installed.

Instagram should be running in a sandbox that gets truth only if the user actually wishes to give it the truth, and if they want to demand unnecessary info from the user (e.g. access to data about other installed apps), then the app should be either lied to by the OS or banned by the platform provider.

There's an inherent conflict between the needs of the app developers and users. The device manufacturers have to take a side in that conflict, presumably the side of the user who's paying them for the device.

I really like this approach

> then the app should be either lied to by the OS or banned by the platform provider

The problem with it is that at least one of the big the platform providers is in a conflict of interest, given how they engage in the exact same practices as a business model. Google cannot hit at anyone for any practice they also use without shooting themselves in the foot. So the status quo is maintained with everyone being allowed to take a swing at vacuuming as much personal user data as humanly possible without ever informing the user or getting proper (informed) consent.

I'm not getting a phone from Google, I'm paying money to Samsung or Motorola or Huawei or Sony or whatever who all already control and customize the OS installed on the phone; have an incentive to compete with other somewhat interchangeable manufacturers by being user-friendly; and don't have any vested interest in third party app vendors being able to obtain and sell data.

LineageOS offers this through Privacy Guard. But it's been pushed deeper and deeper in the settings hierarchy, and will be removed next release due to maintenance burden.

I block ads and trackers where I can but also recognize the value in understanding how one's app is used.

So I'm curious how privacy-focused mods figure out which features users most want without being able to track users. It seems like a huge risk to remove a feature that could be your killer feature but you just don't know.

How did product manufacturers figure out what features users wanted most before modern technology? I'm guessing they didn't just set up a hidden camera in their homes and spy on them using it without their permission.

They asked customers, and developed based on feedback from engineers, PMs, etc. As someone who works in an industry where people are vocal about what we should be doing, what people do and what people say they do are two different things. That, and products are different now. Software releases are daily if not more often, and to keep up with that sort of pace, you need data quicker than you can get in and analyze hundreds of people.

There must be a middle ground though. It can't be necessary for me to share my location data with an app for analytics, nor should it be necessary for me to provide a timestamped personally identifiavle breadcrumb trail, just so a site can know how i navigated.

I noticed the quality of software has declined significantly in the last decade despite more and more telemetry & analytics being used. I'm not convinced analytics provide more valuable insights than committed customers' feedback.

They're answering two different questions: what do vocal customers say they want and what does the mean customer actually do?

History is littered with beautiful products that only like five people actually wanted. I love the Chevy Volt, it fits my use case perfectly and is everything I would ask for in a hybrid right now, but it's sales numbers were objectively crap. It was basically a Chevy Cruze, but sold something like 1/20th the number of units over the same time period; I'm not surprised it was discontinued.

Analytics-driven design leads to unexciting products that are useful enough to most people, with only occasional forays outside the mean when a subteam manages to launch an "unnecessary" feature with neutral metrics (until another subteam discovered the feature can be turned down with neutral metrics). It's not surprising companies love that style of development.

> They're answering two different questions: what do vocal customers say they want and what does the mean customer actually do?

The only question they're answering is "how do we maximize our profit?". If that happens to mean adding features users want that's just a nice side effect.

Claiming companies only care about profit is a bad faith argument. By that argument, the only reason the product exists in the first place is for profit. It's coming from the same place as claiming the only reason people work for companies is because they would starve to death otherwise. Distilling the argument to that point removes any opportunity to discuss the actual question being asked.

I work on a live service, and we use analytics to determine the impact of a crash. If someonr crsshes, often they go straight to reddit and conplain If you took everything at face value online, you would never correctly assess the impact of an issue, and you'd spend all your time fixing rounding errors of issues rather than improving the experience for a large number of people. If you ask people, they'll tell you it's the worst issue, even if it's never happened to them.

How does the solution proposed by your parent prevent software companies from tracking what features their users are using? App developers don't need to be able to track what other apps are installed or most of the other PII that these ad platforms track to determine what features are being used. Grindr could easily tell what features a user is using via logging what the user requests from their servers; after all a social dating app like that doesn't really have many features that don't rely on a network connection. So blocking the additional tracking wouldn't prevent them from seeing what information the user pulls/sends to their server and thus what actions the user is taking in the app. The detail might not be as fine, but I'm less than convinced that knowing the millisecond timing a user takes to perform an action is really all that helpful anyway.

Use statistics from those users who give intentional, explicit opt-in consent.

The problem is distribution. You can’t get even 5% of people to actually care enough about privacy to make a configuration change.

That shouldn’t be the responsibility of the users. It’s the responsibility of the manufacturer, even if the end user isn’t aware of or care for those improvements.

This takes place in every industry except for software development.

> It’s the responsibility of the manufacturer, even if the end user isn’t aware of or care for those improvements.

Manufacturers are incentivized to install bloatware and, in some cases, outright malware.

Fixing this requires legal action.

> It’s the responsibility of the manufacturer

In case of google (android), it's the manufacturer (of the OS) earning money on this

This is an amazing idea.

I totally agree. Beyond the obvious privacy gains (which I am very much here for) I'd love to be able to take on different virtual personas within some of these apps just to understand the media landscape of people other than myself. It's always a bit of a trip when to see how the web experience changes when I borrow my girlfriend's laptop, and she's pretty similar to me in the great scheme of things.

I'm not an app developer. Would this be technically doable on Android? Isn't there already support for running certain apps in a sandbox/container?

I _think_ cyanogenmod exposes some of this functionality.

But as soon as it’s implemented, snoopy apps will just raise the bar further (sms verification, face verification, ‘add this contact to your phone book’, etc)

They can verify SMS all they want but they'll sell my location data consisting only of my random floats. I don't mind legitimate uses, like verifying you're a human or a real user for that matter.

Yep. It's going to be an arms race. So be it. Our AI vs. adtech AI.

It may just be that once this exists then you end up in the "software engineer who uses hn " bucket and will just see adds for ssds and processors and coding conferences... A segment that is really hard to identify may turn into a group that is generally easy to identify...

That's good, that's how we know it works. From there, it can spread to family members, friends, etc.

Please don’t. I can’t beat google captchas as it is.

It has been done by XPrivacy and XPrivacyLua.

Think of these apps as malware,after all they are hostile. In dynamic malware analysis, virtualized and emulated environments are used to run malware samples to analyze their behavior. It's like a tom and jerry cat and mouse game with the jerry always winning (like the show!).

It's trivial to detect virtualized environments , not only that, the samples can afford to be patient and meticulous,and test themselves against existing analysis methods before deployment. That's why real hardware is the best place to study malware.

Your disadvantage here is that fingerprinting is a game of accuracy, where the adversary gains accuracy at basically a logarithmic(or geometric?) scale for every additional data point they have about your device, sometimes they only need one data point,such as a shared file system,let me explain: most apps need storage access right? Like you want them to read existing files or write files to be accesses by other apps. App1 writes a file with a random name or random content, app2 through appN in their affiliate network check that location/file for the random value to track your device. The apps in this network all collect various data points about you, let's say one app gets your real name and address (payment and shipping),now all the apps know those two things about you. Now they all know your zipcode and can infer things about you based on that,your last name,etc...and even do background check like lookups to know your past addresses aquaintances, cars you've driven,court cases and so much more. Now let's say one of these apps needs to send you text before it can function(most popular apps these days) , now all of them know your phone,can get a ton of info about you,who else has you as contacts and they can infer even more about if at least one of the people they track has your number as contact and they basically opened up their lives to all apps.

You can add all the layers of abstraction you want. Apps need to talk to each other and external sites/services need to reach you at a predictabl address (including physical address and such). And even if what you said was effective, it's incredibly hard to achieve without the user being very cooperative towards their own privacy preservation, and evem then ,like I said above, if anyone you have as contact reaches you using a predictable address, a lot can be constructed about solely based on their choices.

My solution is legal. Make it a crime and start throwing people in prison. If someone violates your consent to stalk you or creates public opportunities of business or employment where consenting to stalking is a requirement, make it a felony. And make privacy a core right for peoples of any society that calls itself free.

Unlike with malware authors, we can easily find the execs behind stalkerware and we can make them take responsibility (assuming people want liberty and democracy more than political theatrics).

I agree with you. But as an engineer, it's easier for me to envision technical solutions than political ones. Both are probably essential.

Doesn't apple with it's ever tightening restrictions of what data you can ask for, like no more MAC address, no more device UUID other than an apple provided "ad id" that can be reset, along with 2 bits you store with apple, aggressive always on location nag alerts, etc also accomplish something similar bit by bit?

But there's always the issue that the app might need some data for functionality but also misuses it. Like having you type in notes and syncs it with cloud, but then uses the data to show adverts.

Facebook is still in the store. Instagram is still in the store. Grindr (the app in the article) is still in the store.

Facebook Messenger actually has a nasty dark pattern where they will display a fake pixel-perfect copy of a permission prompt where the only way to get through is to tap Allow, and when you do they call the real prompt (since it's pixel-perfect you can't tell the difference besides the short animation when it appears) and hope you also click Allow on the real one. If this is allowed then it's a lost cause.

Apple is doing fuck all about privacy on the App Store.

Brb going to test this wild claim

Edit: nope.

Nope as in they don't do it anymore? I'd be very curious to see what's the onboarding flow now.

Or an app that provides a link to a remotely hosted desktop or mobile environment to surf / run apps from.