[imdoor]

I'm curious, shouldn't the "charge only" mode, that's the default, when connecting usb stuff to Android phones, be enough to protect users? Is it really that difficult to implement a "don't read data pins, only charge" mode on a phone and not have vulnerabilities in it?

[HeavyFeather]

If you can connect your turned off phone to your computer and start a reset, then that’s never going to be enough.

If you want data safety, you must skip the data pins.

If you want current safety, you must skip public chargers.

[tshaddox]

If it’s “just a reset” I still wouldn’t be too worried plugging into an otherwise normally placed public charger. It would obviously suck to have my device reset, especially when traveling, but of course a port could also just fry your device anyway.

[yencabulator]

If it's just a USB-initiated factory reset, that's much less worrying, just DoS not infiltration. Exploiting that at a busy airport would be a huge nuisance, but not a huge security risk. Just like wiring 110VAC into the USB wires would be a DoS...

[paulsutter]

I would still prefer a “never trust” mode, even if it meant I had to go to an Apple store to do a reset (something I have never needed to do)

[epups]

I don't get it, even after I reset my phone it's still locked, and by default not sharing data via USB. What am I missing?

[retrac]

> not sharing data via USB

USB is a very intelligent protocol, with a microcontrollor on both ends. The controller has access to at least the driver's state, which is usually in the kernel and potentially has access to system memory.

How does your Android phone even know that data is an option to switch into when you plug it into a USB port? It has already negotiated itself to be a device on the USB bus. Your phone will probably show up in lsusb on Linux even in charging mode. (Mine does.) When you switch the phone to data mode, it changes its USB device profile, and becomes a more sophisticated attached device, from the host's perspective.

Many (most?) phones made in recent years can be USB hosts, too. This lets you connect a USB mouse and keyboard to a tablet, for example. That would open you up to all kinds of pretty simple but often quite effective attacks, like simulating a virtual keyboard and mouse and just manipulating the UI that way.

I don't know if any of these particular attacks are possible with Android right now, but many variations on these themes have been shown over the years on many platforms. USB wasn't really designed with adversarial peripherals in mind.

[epups]

Maybe I'm stupid but what I gather from this is simply that this is a potential vector, not that it is currently an actual possibility. It's akin to saying using Bluetooth is dangerous because theoretically any data on my phone can be extracted through it, while neglecting the fact that the people building a phone OS are clearly aware of that and have built-in countermeasures.

[kccqzy]

If the USB connection truly doesn't get data, your charging experience is unsatisfactory: there's no way for the phone to negotiate higher wattage.

Not "sharing data" doesn't really mean not sharing data.

[StrangeATractor]

https://en.wikipedia.org/wiki/BadUSB

[upofadown]

BadUSB emulates a keyboard. So one would want to make sure that the phone was locked before hooking it up to a random charging port. Android exploit demo here:

* https://github.com/caioau/badUSB-Targeting-Android

[rhplus]

You phone can only figure out if it’s connected to a known device (your car, your speaker, etc) by asking the data pins. A charge-only mode would “break” usability of the USB port for most users.

[was8309]

android 11 asks me if i want to charge only or also allow data transfer. Is it that we can't trust android to be not be hacked just by checking if data pins exist?

[ensignavenger]

My phone asks me if I am connected to a trusted device and want to share data, asking me rather than asking the device if it is trusted seems to be an effective model.