Ok, guys. Unfortunately I forced to admit that I have no more time to work on PS3 stuff because I'm very busy lately. So I decided to publish all information related to PS2 classics as @JuanNadie did with the NPDRM algorithm one year ago. Firstly I wanted to say that he was the first who started reverse-engineering on this subject and when he left the scene I decided to continue his work to keep it from going to waste. And so I would like to thank @JuanNadie for his amazing contribution to the PS3 scene. Besides that, he gave me some piece of information on the subject.
All PS2 classics runs within the
ps2_netemu.self which represents a different kernel for execution these PS2 games but before it started the VSH module loads your individual data for PSN/SEN (such as act.dat and .rif file for your game). It is absolutely the same process as used for usual PSN games and the goal of it is getting the key used for decryption of PS2 content which includes an optional CONFIG file, ISO.BIN.EDAT and ISO.BIN.ENC. The latest one is the actual encrypted disc image of the game. All mentioned files are encrypted with the same key (called klicensee) which is stored in encrypted form inside .rif file for your game and it decrypted with the specified key from key table stored in act.dat. When you get this key you can decrypt ISO.BIN.EDAT and see if it contains a game title (for example, SLUS-20062 for GTA 3). This will mean that key is correct. Since almost all the information regarding EDATs is known (see there and there) I will not going to explain it again.
Well, now there are two another formats along with
EDAT. Let's call the first one as
ENC (it represents the actual disc image) and the second as
VME (encrypted virtual memory cards). They are encrypted using different algorithms. The ENC format is similar to EDAT and the VME format have a simple encryption layer.
As I said before, ENC file is similar to EDAT and it have the header like in EDAT (but with different magic) and composed of segments of 16384 bytes each (you can see it at the header). I just remind you that file header consists of file magic (PS2\x00), version number (major and minor: 01.01), license type (it always 0x02), application type (0x01), content id, QA digest (seems like to be a SHA-1 hash of the non-finalized file generated using the tool from SDK), CID-FN hash (an AES CMAC hash of concatenation of content id and file name using the third NPDRM OMAC key as CMAC key), header hash (an AES CMAC hash of the 0x60 bytes from the beginning of file using xored bytes of the first NPDRM OMAC key and the second NPDRM OMAC key as CMAC key), time information which includes start and end time of the validity period (they are usually zeroed, base ticks = 62135596800000000), file flags (always zeros), segment size (16384 bytes), data size of the file data, two unknown hashes of 16 bytes each, 40 bytes of unknown data (possible another unknown signature) and pair of an ECDSA signature (40 bytes using the second VSH curve and the VSH public key). I also remind you that two unknown hashes for EDAT case are known and represents meta data sections hash and extended header hash (an AES CMAC hash of 160 bytes from the beginning of file), both hashes uses the hash key as CMAC key and it depends on the file flags and keys). I don't know exactly what hashes are there for ENC format but when we zeroed them it seems like they are not checked on current firmwares. The file header ends at the offset of 256 bytes.
Segments are divided into two types: a meta data section and a file data section. Each meta data section can include 512 entries (max) of 32 bytes each (16384 / 32 = 512) and associates with a particular file data section. So if we have a meta data section which consists of 512 entries then it will mean that there are 512 file data sections after it and each file data section have size of 16384 bytes. Besides that, the first meta data segment located at the offset of 16384 bytes. I don't know what data are stored before it but we also tried to zero them (these bytes starting at the offset of 256 bytes and ending at the offset of 16384 bytes) and it works as usual. I guess that it can be the encrypted garbage because the alignment of file data should be equal to the segment size.
Now I will explain what keys are used and how they are obtained. ENC/VME files are decrypted using the ENCDEC device so the decryption process are more faster than at EDAT case. While vSH checks files for their validity period, CMAC hashes and ECDSA signature and obtains the key for decryption from .rif file and it makes a system call #475 to LV2 (on older firmwares it was #471) along with the NPDRM information, klicensee, act.dat key and encrypted rif key. LV2 gets your console ID, encrypts the NPDRM constant using it as a key, decrypts the key from act.dat using the encrypted NPDRM constant and finally decrypts klicensee from .rif using the decrypted key from act.dat. Now we have a klicensee which will be used for later decryption process. For EDAT case we can use free EDATs without .rif but for PS2 classics we should always use paid content and .rif file. So if you want to resign the game you need to generate .rif for the account on your console (I call this process as "personalization"). Don't forget that .rif file should be created for your act.dat (because it shares the account id) and console ID. Let's move on. When the PS3 gets the final decryption key it send a packet to the system manager inside LV1 which sets the inter-lpar parameter of type 3. This parameter contains a version information and the klicensee. A system manager catches this packet and sends a request to the storage manager inside SS server #1 which then configures ENCDEC keys used for later decryption. It should be kept in mind that keys for decryption differs between CEX and DEX consoles so the storage manager checks the device type and uses different key slots for ENCDEC. The configuration process started with running isolated SPU SB module which creates the final keys using klicensee as a key seed and send them back to the PPU which then send them to the device directly during the secure session. There are three types of keys: meta key, data key and vmc key and they are configured separately. The process of making keys consists of applying an AES 128 algorithm on the klicensee while using three different keys.
There are SHA-1 hashes of each of three keys (you should decrypt sb_iso_spu_module.self from 4.xx FW and find each of 16 bytes key by its SHA-1 hash):
For CEX mode:
Code:
1. Meta key: B9CACFF9E126F63634DC38AF61040BDF6F370A26
2. Data key: CB0BAECAAADF9E5C629522B11757F78C7CD5B23C
3. VMC key: EB03D83F96E3394A05BCE68F8645DA134CDA5545
For DEX mode (you actually don't need it but anyways):
Code:
1. Meta key: 4FCFB6683AC46E73FFFCE49895E3F303A117BE8C
2. Data key: AEC7A9C13A4023FE268A163FFDC8382F45496928
3. VMC key: B41AEE9D3B6C54292469C9C754AE8FE75ACBE958
Now we have all keys which are required to decrypt all files. So what we should also know?
ENC encryption uses an AES algorithm in CBC mode and the initialization vector of all zeros. The actual process of decryption of CONFIG and ISO.BIN.ENC started at seeking to the offset of 16384 bytes. There is a first meta data section so we should use the meta key as key for AES and decrypt the entire segment of 16384 bytes. As I said before each meta data sections contains of some entries and each entry have a size of 32 bytes. Each entry contains a SHA-1 hash (20 bytes) of the corresponding entire encrypted file data section and all these sections are located after this meta data section. After the SHA-1 hash we can see the section index of the corresponding file data section (4 bytes). The rest is padded of zeros. After decryption of the meta data section we can decrypt all file data sections after it. Now we should use the data key! Before the actual decryption we can check the SHA-1 hash of each encrypted file data section and see if they matched to the hashes at entry table of the meta data section. If the actual file size of the disc image is not a multiple of 16834 bytes then we have less entries inside the latest meta data section. After we finished the decryption of first 512 file data sections we can started decryption of the second meta data section and set of 512 file data sections after it and so on. I recommend to write decrypted meta data entries to another file than in the same file as file data section. It will make a process more easier. After decryption you should truncate your actual file to the data size specified at the header. Now you got an UDF disc image and you can mount it on your PC, for example.
So what is the next step? The next step is the decryption of encrypted virtual memory cards. Each PS2 classics package contains two empty encrypted virtual memory cards which located at SCEVMC0.VME and SCEVMC1.VME. As far I see they are identical for all games so we can use templates for all new virtual memory cards but only encrypts them with the new klicensee. To decrypt virtual memory cards you need to read an each segment of 16384 bytes and apply an AES encryption in CBC mode too but for this case you should use the VMC key. After decryption you should see Sony PS2 Memory Card Format 1.2.0.0 at the top of file.
Well, I attached a draft script for decryption of ENC/VME files. It was written for Python 2.7 and requires CryptoPlus (can be downloaded from:
http://repo.or.cz/w/python-cryptoplus.git) and "ecdsa" (use EasyInstall or another package manager) libraries. I intentionally left all keys as SHA-1 hashes because of legal issues but you can find all keys by yourself using my hints. My script uses CONFIG/ISO.BIN.ENC/SCEVM0.VME/SCEVM1.VME file and klicensee file as input parameters. I hope that someone will create tools for that.
To use the script you need to create a file with name vsh.curves and put the contents of the curve table from VSH (get it from
http://ps3devwiki.com/wiki/Keys at vsh pub + curvetable) and replace all hashes of keys by their real values (see FIXME comments). Also replace three NPDRM OMAC keys and VSH public key by their values from
http://ps3devwiki.com/wiki/Keys.
I think that creation of PS2 remastering tool can lead us to getting the fully working games on our consoles but it requires testing. I recommend to create a static klicensee which can be used to encrypt all images in the same manner (static klicensee can also be implemented by patching VSH/LV2 at runtime, for example). After generating a klicensee you should create all keys based on it.
To build an encrypted disc image you should dump the original disc image and then append zero bytes to the end to make it multiple of 16384 bytes. Then you need to encrypt each of 512 segments using the generated data key. Then you should calculate SHA-1 hashes of each encrypted segment and generate meta data section for each pair of segment hash and segment index. After this you need to encrypt meta data section and so on. At the end you need to write an original disc image size to the header, write a content id for it and generate hashes at the file header.
After building ISO.BIN.ENC file you should create a file with the title id and pad it with zero bytes from the right side to get 12 bytes total. Then you need to create an EDAT container for this file. Hint: you can see a correct title id when mounting a disc image on your PC and looking at SYSTEM.CNF of it.
Unfortunately, I hadn't time to see what the CONFIG file does so I will skip this step. I only know that this file is optional or can be empty inside (after decryption).
You are not required (and you simply can't do it) to generate a valid ECDSA signature for files because all custom firmwares are patched to skip the ECDSA check.
Will be nice to be able to generate a game package for your PS2 game too if everything will works fine. Remember, that some flags at PS2 pkg format can be different.
http://pastie.org/private/ykyyrbim7rkotvcnw8m8w
Credits to:
graf_chokolo, fail0verflow, JuanNadie, ps3dev.net, glevand and all my friends (you know who you are).
1 Reply
|
422 Views
EXCLUSIVE: Team REBUG updates TOOLBOX and REX firmware
'PS2 Classics Algorithm' published by flatz
Nvidia: 'PS4 not worth the cost'
Originally Posted by Cyberskunk
