September 15, 2013

Review of USB Condoms

We're in danger

Computer software anomalies often have their biological analogies such as viruses or bugs. Transmission of these virtual infectious diseases by physical contact is observed since the dawn of the digital age, be it on floppy disks, optical disks or USB drives. But recently, a new trend of threat transmission targeting mobile systems began to spread. Since the mobile era, more and more devices (cellphones, music players, etc.) use their USB connector for multiple purposes such as data transfer and charge. This allows malicious computer programs to inject malware on a device when its owner only wants to charge it. Worst: one can modify a simple charger to propagate these treats. The problem? It begins to spread to public and mainstream systems.

The solution to these virtual physically transmitted diseases? An USB Condom! Available through, this simple board only connects the power pins through while cutting the data pins. Simple and clever. But can we presumptuously remove half the USB 2.0 wires without side effects?

The current state of affairs

The USB connectors are defined with 4 (2.0) or 10 (3.0) connections which always contain two power pins (5V and ground) in addition to their data pins.

Let's read some power specifications directly from the ultimate reference, the Kamasutra of USB, the official USB 2.0 standard documentation, pages 171 and 245:
Depending on the power capabilities of the port to which the device is attached, a USB device may be able to draw up to five unit loads from V BUS after configuration. A unit load is defined to be 100 mA.
It is also written on page 178 of the USB 2.0 specifications document that a device described as high-power should require a maximum of 500 mA on the 5 V power rail of the USB connector after being configured and 100 mA maximum during power-up (page 171).

The USB 3.1 specification rise the high-power current draw to a maximum of 900 mA (6 unit loads of 150 mA) when the client device is configured but states, on chapter 11.4 of its specifications:
Note that a USB 3.1 peripheral device shall not draw more than 100 mA until it detects far-end Rx terminations in the unconfigured state.
Finally, - the last quote from the specifications, I promise - the battery charging specification allows (in its revision 1.2 from December 2010) to draw up to a whooping 1500 mA from an USB port of a compliant device for charging purposes without the need to connect the data pins (after a delay of maximum 900 ms). But this amazing power is only available through hardware modifications of the original USB host specification, which most computers and laptops don't implement.

It should be noted that many manufacturers of portable devices and adapters do not follow these specifications. Most wall adapters can deliver more power than any standard specification, even the battery charging specification. For example, the ODROID-U2 uses an EIAJ-01-like connector to transmit 10 Watts! But who will condemn this practice - aside the firefighters - when it means faster charge times?

With this kind of devices, the USB Condom will do a good job protecting your beloved electronics from electronic STDs while allowing them to charge freely.

But you won't have this luck when charging from a computer, a laptop or any other power delivering device that is more likely to follow specifications. The USB enumeration process, which uses the data pins, will likely fail when using the USB Condom. The device providing power will not be able to read the configuration registers stating the high power requirement of the charging device. This will produce a maximum charge of 100 mA, which is the maximum allowed current during power-up (unconfigured state) and which coincidentally can be near or under the device power consumption. Result? An awfully slow charge or even no charge at all!

This behavior particularly hinders the utility of the USB Condom since its goal is to charge your portable devices while protecting it from malicious communications which are mainly from computers - which may have viruses - or a fraudulent charger containing an embedded microcontroller delivering malwares. Using the USB Condom will result in a slow or no charge from these sources, depending on how much the pirate manufacturer of the charger respected the specifications. It's a shame; when you put a condom, it should draw some satisfaction!

There shall be light

One way to circumvent this problem would be to include a USB charger detector IC or a simple microcontroller with an USB stack such as the ones available from Microchip. While keeping the power connectors linked to the protected device, its data pins would be replaced by the microcontroller's pins that would configure as a generic high-power device while ignoring any communication attempt. Sadly, this will boost the manufacturing price of the USB Condoms. But it will allow the charging sequence to respect the original USB specifications. This new device could even capture every malware injection attempt and create an Virtual STD Honeypot! What an appetizing idea!