Saturday, December 14, 2019

iOS Imgur app - A Realm database example

Short version:

The Imgur app keeps user generated post data in the following Realm database:
/*app folder GUID*/Documents/default.realm
All images accessed via the app, from images posted by the user to those seen as one navigates through the content the app presents, are found in the following location:
 /*app folder GUID*/Library/Caches/com.hackemist.SDImageCache/default
 App screen snapshot can be found at:
/*app folder GUID*/Library/Caches/Snapshots/imgurmobile/*.@3x.ktx
For details on Real databases and how to approach their examinations see here:
It is of note that Cellebrite Physical Analyzer has a database browser that is compatible with Realm databases.

For details on iOS snapshots and how to visualize them see here:
It is of note that Magnet Axiom has the capability of displaying .ktx images natively.

For details on how to access a full file system of an iOS device see following:
  1. Checkra1n jailbreak:
  2. Install OpenSSH via Cydia. See the acqusition section for SSH here:
I will update the post with a Checkra1n step-by-step guide that is being prepared by colleage soon.
It is of note that Magnet Acquire, a free tool, allows the creation of a full file system .tar file from a Checkra1n jailbroken device.

Long version

The Imgur iOS app is one of the growing number of software development project that have moved from using SQLite databases to the Real database NoSQL data storage type. For details on how to approach Realm databases see my previous post on the topic here:

This post is not an exhaustive study of the Imgur iOS app. I use it as an example of the type of app that is used by millions of people that might come across your case work and your tools might not have the default capabilities to address at this moment.

In order to locate a particular application folder, in our case the one for Imgur, within our full file system extraction see Nike iOS app example here:

After locating the application directory we can exported from the device or from the tar file for analysis. So far these are the steps:

Jailbreak the device using Checkra1n.

macOS only for now.

Install OpenSSH via Cydia.

Open the checkra1n app and install Cydia. then open Cydia and install OpenSSH.
Locate the applicationstate.db file in order to identify where the Imgur user generated data resides. See previous links for steps required.

Connected from macOS to the phone via SSH. Cyberduck app. Username root and password alpine.

Export the application directory.

To figure out which one you want see the applicationstate.db entry for the app you want.

Open the application directory for analysis.

The bottom window shows the application directory folders.

Imgur analysis

The Realm database we are looking for will be located in the Documents directory. As seen in the next image my macOS computer identifies the datastore with a red and bluish icon. This is due to having already installed Realm Studio. This application allows us to access the contents of Realm databases. It can be downloaded here:

Get the default.real file.
Open the file with Realm Studio.

The data seen in the previous image corresponds to me making a post to Imgur using the app. The post includes an image. The left column identifies the objects and the right pane has the corresponding properties. This database separates the data in the post form the image that goes in the post. The image post has an entry in the DraftImage object at the top. This screen shows the post tile and some metadata about the post.


The DrafPost object contains two important properties, the remoteidentifieer and the accounintidentifier. The remoteidentifier is the post number used by Imgur in the website URL. The accountindentifier is self explanatory. It is tied to the user.

The DraftImage object has the name of the image in the remoteidentifier field. To confirm one can go to the post and right-click save the image. The name provided will be the same as the one found in the database. The DraftImage object also contains additional metadata as image width and height.

As you can imagine the analysis of a database made of object can get really difficult pretty fast as more objects are created and stored. One technique is to export the data as a JSON file and use Python or other tools to create a report of the contents.

Export file to JSON.
The following is a screenshot of the output of my JSON-to-HTML Python script for the exported JSON data.

Easier to visualize
I find it easier to visualize the data in a nested columns and rows format. The script can be found here:

One last artifact I wanted to highlight are the application snapshots. These are screenshots of what the app was showing on screen when the user places the app in the background in order to have a different app in the foreground. Be aware that the ktx snapshot is not compatible with regular image viewers outside or macOS and Magnet Axiom.

For details on snapshots see @iam_the_gia's blog post here:

For details on how to identify ktx files and how to convert them to png see the blog post here:


By generating more test data we can surely identify additional objects and artifacts of interest. The main takeaway from this exercise is the existence of app datastores that go beyond the usual SQLite we are used to. The same thing can be said of JSON and protobuff data stores. The only permanent thing in the digital forensics field is the fact that there will always be change. It is upon us to discover, test, and publicize new changes. Only then our tools will be updated, evidence won't be overlooked, and justice will be served.

As always remember to validate all findings and be aware I can be reached on twitter @AlexisBrignoni and via email 4n6[at]abrignoni[dot]com.