This repository is an educational (not production-ready) class project to mount and log writes to a remote block device on Android. The logging is done with the GammaRay project, which provides filesystem-level streams of block device operations.

Corporate phones can use a network block device with LVM to mirror Android's physical block devices as files on a remote server to synchronize the device containing /system to a central server when employees are on-site.

The amount of applications on the marketplace is approaching 1 million, and seemingly benign applications could perform malicious actions on a user's device. This use case would involve using nbd with LVM to mirror the phone's block devices to a remote server that has the capabilities of streaming the block writes in real time when an application is performing.

Then, each application can be experimentally profiled and paired with a sequence of hard disk reads and writes. These sequences could reveal the nature of the application, and perhaps a malicious application could somehow gain privilege escalation and modify the contents of /system.

This technique of malware analysis by associating an application with some stream of events is called dynamic analysis, and the paper [Applying machine learning classifiers to dynamic Android malware detection at scale][antimalware] further discusses this approach. [antimalware]: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6583806

Consider a systems-level developer modifying the Android system and Linux kernel. Some change that she made to a complex internal function call is causing kernel panics and thinks monitoring the state of the device's storage leading up to the crash will help reveal the bug. Low level filesystem debugging can be done with nbd with LVM to mirror and stream the hard disk writes to a remote server.

GammaRay provides disk-based introspection for virtual machines as shown in the following diagram. The block device writes are intercepted and Redis maintains the streaming metadata. The inference engine infers the filesystem operations that occurred in some time interval.

This repository uses Network Block Device (nbd) on Android to connect to an export provided by GammaRay as shown in the following diagram. Writes to the nbd on Android are stored in Redis on the server and can be introspected.

Install and run redis-server and build the GammaRay binaries. Run demo/init-server.sh to create the nbd export of the partition on port 30004 through GammaRay's custom nbd queuer. This will use Python to start an HTTP server on port 8000 to view the contents of the block device.

Next, follow the instructions in this blog post to connect to the server's nbd export on port 30004. Mount the partition to /sdcard/DCIM/Camera and pictures taken from the device's camera will be captured on the GammaRay server.