Decrypting Firmwares

iOS contains many layers of encryption. This page details how to remove the encryption wrapper around each file in the IPSW file.

Ramdisks

This section details the decryption of the ramdisks in an IPSW file. The listed console commands are applicable to the IMG2 or IMG3 files under /Firmware also.

1.0.x

With the release of the iPhone, the ramdisks weren't encrypted. So, in order to mount them, all you need to do is remove some data from the beginning. You can either open up a hex editor and remove 2048 bytes (2 KiB) from the beginning, or open up a console and run dd(1)^[man]:

dd if=ramdisk.dmg of=ramdisk.stripped.dmg bs=512 skip=4 conv=sync

where ramdisk.dmg is the filename of the restore ramdisk (ex: the iPhone 1.0 firmware (1A543a) would be 694-5259-38.dmg)

where ramdisk.stripped.dmg is the output file name

Once the data has been stripped, you can then mount ramdisk.stripped.dmg in Finder on OS X, or with any other program. If you encounter errors after mounting the stripped ramdisk, you can safely ignore them.

1.1.x - 2.0b3

With the release of the iPod touch, Apple added a layer of encryption around the ramdisks. The decryption key wasn't obscured however, and a simple analysis of iBoot by Zibri revealed the 0x837 key. At first, its purpose wasn't known. After a while, geohot discovered its purpose.

In order to decrypt them, all you need to do is remove the 0x800 byte (2 kibibytes) header, then open a console and run openssl(1)^[man]:

openssl enc -d -in ramdisk.dmg -out ramdisk.decrypted.dmg -aes-128-cbc -K 188458a6d15034dfe386f23b61d43774 -iv 0

where ramdisk.dmg is the filename of the ramdisk you are decrypting (ex: the iPhone 1.1.1 firmware (3A109a) would be either 022-3630-9.dmg or 022-3629-9.dmg)

where ramdisk.decrypted.dmg is the output file name

2.0b4 - 3.0b5

With the fourth beta of 2.0, Apple introduced the IMG3 file format, replacing the broken IMG2 file format. This format was soon reversed and img3decrypt^[src] was created by Steven Smith (@stroughtonsmith). His code was later implemented into xpwntool^[src]. In order to decrypt the ramdisk, open a console and run one of the commands depending on your program choice:

img3decrypt e ramdisk.dmg ramdisk.decrypted.dmg iv key
xpwntool ramdisk.dmg ramdisk.decrypted.dmg -k key -iv iv

where ramdisk.dmg is the filename of the ramdisk you are decrypting (ex: the iPhone 3G 2.0 firmware (5A347) would be 018-3783-2.dmg)

where ramdisk.decrypted.dmg is the output file name

where iv is the initialization vector (IV) of the ramdisk you are decrypting (ex: the iPhone 3G 2.0 firmware (5A347) would be 29681f625d1f61271ec3116601b8bcde)

where key is the key of the ramdisk you are decrypting (ex: the iPhone 3G 2.0 firmware (5A347) would be 850afc271132d15ae6989565567e65bf)

The IV and key for a specified firmware is available through the Firmware Keys page or from the Info.plist file underneath PwnageTool's /FirmwareBundles folder.

3.0GM/3.0

OS X Snow Leopard introduced the HFS compressed disk image. With 3.0 (what beta?), Apple began using Snow Leopard to package the ramdisks. This results in some zero sized files in the disk image if you don't use Snow Leopard or newer. A discussion on extracting those files is available on the talk page.

S5L8900

With the 3.0 Golden Master (7A341), Apple messed up and, instead of using the application processor-specific GID Key, used a pseudo-GID of 5f650295e1fffc97ce77abd49dd955b3 to encrypt the KBAG. This makes obtaining the keys for this version dead simple. Once you have decrypted the KBAG, decryption using the keys in it is the same as above.

S5L8720

Business as usual, but keys and IVs have to be decrypted on the device still, unlike with the new S5L8900 KBAGs. Apple incorrectly assumed that by encrypting iBEC and iBSS they were being sly. They were not. You can decrypt those on a 2.2.1 aes setup no problem whatsoever.

S5L8920

The iPhone 3GS firmware files are interesting. They have two KBAGs, which use AES-256 instead of the S5L8900 and S5L8720 that are using AES-128 still. The first KBAG has an identifier in it's header indicating that it is to be decrypted with the gid key, and the second is not known. For those that don't know how AES256 works, this now means that the first 0x10 bytes are the IV, and the remaining 0x20 bytes (not 0x10 anymore!) are the key.