Talk:SELF - SPRX
To be merged with SELF File Format and Decryption
NPDRM Header
typedef struct
{
u32 block_type; // this is 3(NPDRM)
u32 block_size; // this is 0x90(sizeof(Self_NPDRM))
u32 unknown1; //So far always 0
u32 unknown2; //So far always 0
u32 magic; // 0x4E504400(NPD)
u32 unknown3; // So far always 1
u32 license; // 1 Network License, 2 Local License, 3 Free
u32 type; // 1 Executable, 21 Update for Disc Based Game
u8 titleid[0x30];
u8 hash_unknown[0x10];
u8 hash1[0x10];
u8 hash2[0x10];
u8 padding[0x10];
} Self_NPDRM
Located after the Self Control Info.
App Info header:
Aligned to 0x10 bytes.
| field | offset | type | notes |
|---|---|---|---|
| authid | 0x00 | u64 | |
| unknown | 0x08 | u32 | |
| app_type | 0x0c | u32 |
|
| app_version | 0x10 | u64 |
Encrypted phdr offset entry
There is one of these entries for each phdr entry in the elf file so that the ps3 knows where to decrypt the data from. (because it might also be compressed.)
| field | offset | type | notes |
|---|---|---|---|
| Encrypted Data Offset | 0x00 | u64 | |
| Encrypted Data Size | 0x08 | u64 | |
| unknown | 0x10 | u32 | This has been 1 in all the examples I have seen. |
| unknown | 0x14 | u32 | Always 0, as far as I know. |
| unknown | 0x18 | u32 | Always 0, as far as I know. |
| unknown | 0x1c | u32 | This is 2 for loadable segment types, and 0 for other types. |
| field | offset | type | notes |
|---|---|---|---|
| Magic | 0x0 | u32 | Must be "SCE\0" |
| version | 0x4 | u32 | This must be 2 or the Self loader will abort |
| flags | 0x8 | u16 |
0001: FW 1.00 (app version 1.0.0) ... 0001: FW 3.15 (app version 3.15.0) 0001: FW 3.20 (app version 3.20.0) 0001: FW 3.21 (app version 3.21.0) 0001: FW 3.30 (app version 3.30.0) 0004: FW 3.40 (app version 3.40.0) 0004: FW 3.41 (app version 3.40.0) 0004: FW 3.42 (app version 3.40.0) 0007: FW 3.50 (app version 3.55.0) 000a: FW 3.55 (app version 3.55.0) 000d: FW 3.56 (app version 3.56.0) 0010: FW 3.60 (app version 3.60.0) 0010: FW 3.61 (app version 3.61.0) 0013: FW 3.65 (app version 3.65.0) 0013: FW 3.66 (app version 3.66.0) 0016: FW 3.70 (app version 3.70.0) 0016: FW 3.72 (app version 3.70.0) |
NPDRM Self algorithm
THIS DOES NOT ALLOW TO OBTAIN 3.60+ keys, nor piracy as you require the rif, act.dat and IDPS
On NPDRM self decryption all the security levels of the PS3 are involved: user space (vsh), kernel space(lv2), hypervisor( lv1) and isolated SPU (metldr + appldr)
The process start on vsh.elf...
VSH
Once the vsh detects that user is trying to start a self, it looks for the appinfo header type. If the type is 8, then the control digest element type 3 (NPD element) is located. From this NPD header the vsh gets the license type (free, local or network license).
If a free content(type 3) is detected then a generic klicense will be use for further steps (go to LV2). That klicensee is already public (see geohot npdrm_omac_key_1).
npdrm_omac_key1 : 72F990788F9CFF745725F08E4C128387 # ps3publictools/include/oddkeys.h npdrm_omac_key2 : 6BA52976EFDA16EF3C339FB2971E256B # ... npdrm_omac_key3 : 9B515FEACF75064981AA604D91A54E97 # ...
However if a paid content is to be loaded the vsh loads the act.dat and the rif associated to the content (if local it will locate a file with the same titleid on NPD element, if remote it will download to vsh process memory)
Then the signature is checked (last 0x28 bytes of both RIF and act.dat). The curves used are on vsh.self. It is a 3 element table, having the first curve nulled. The curve index for rif/act is 2. The curve values are negated as in the apploader and has the following structure
struct curve {
uint8_t p[0x14];
uint8_t a[0x14];
uint8_t b[0x14];
uint8_t N[0x14];
uint8_t Gx[0x14];
uint8_t Gy[0x14];
}
If the curve checks then vsh will process the rif:
struct rif {
uint8_t unk1[0x10]; //version, license type and user number
uint8_t titleid[0x30]; //Content ID
uint8 padding[0xC]; //Padding for randomness
uint32_t actDatIndex; //Key index on act.dat between 0x00 and 0x7F
uint8 key[0x10]; //encrypted klicensee
uint64_t unk2; //timestamp??
uint64_t unk3; //Always 0
uint8_t rs[0x28];
};
struct ACTDAT {
uint8_t unk1[0x10]; //Version, User number
uint8_t keyTable[0x800]; //Key Table
......
uint8_t signature[0x28];
}
Using the RIF_KEY it will obtain the actdatIndex:
AES_KEY rifKey; int result = AES_set_decrypt_key(RIF_KEY, 0x80, &rifKey); AES_decrypt(&rif->padding, &rif->padding, &rifKey);
And finally having the actDat key index the execution pass to LV2 syscall 471
LV2
Lv2 is accessed using syscall471 which haves the following syntax:
int syscall_471(uint32_t type, char* titleID, void* klicensee, uint8_t* actdat, uint8_t* rif, int32_t licenseType, uint8_t* magicVersion);
The function has different parameters depending if the content is debug, free or paid:
FREE: syscall471(npd.type, &npd.titleID, freeklicensee, NULL, NULL, npd.license, &npd); PAID: syscall471(npd.type, &npd.titleID, NULL, &actdat.keyTable[rif.actDatIndex], &rif.key, npd.license, &npd);
The lv2 keeps a memory table with contentID and the associated key. When it receives a free content (r5 is not null) then copies the titleID and the klicensee to the table. For a paid content the rif.key is converted to the klicensee using:
AES_KEY IDPSKey, ConstKey, ActDatKey; uint8_t encrConst[0x10]; uint8_t decryptedActDat[0x10]; uint8_t klicensee[0x10]; int result = AES_set_encrypt_key(&IDPSVariation, 0x80, &IDPSKey); AES_encrypt(&CONSTACTDAT, &encrConst, &IDPSKey); result = AES_set_decrypt_key(&encrConst,0x80,&ConstKey); AES_decrypt(actDat,&decryptedActDat,&ConstKey); result = AES_set_decrypt_key(&decryptedActDat,0x80,&ActDatKey); AES_decrypt(rif,&klicensee,&ActDatKey);
where CONSTACTDAT is a constant value on lv2, IDPSVaritaion appears to be IDPS (not checked but DRM_Manager_initialize (see graf_chokolo's "bible") to something with the same structure), actdat are the 0x10bytes selected by rif keyIndex, and rif is rif.key (bytes 0x50-0x5f).
Once transformed it is stored on memory table...
I haven't check further steps on vsh nor lv2 so perhaps there are further transformations on the paid case (NOT FOR THE FREE AS I HAVE DECRYPTED THOSE) so we are jumping directly to the appldr
AppLdr
As you can see from graf_chokolo payloads a parameter is passed on spu_args.field60. That parameter is the previously stored klicensee.
However this key must be transformed (again) even for the free case. The transformation is:
uint8_t decryptedKLicensee[0x10]
uint8_t KLicenseeDecryptKey[] = {0xf2, 0xfb, 0xca, 0x7a, 0x75, 0xb0, 0x4e, 0xdc, 0x13, 0x90, 0x63, 0x8c, 0xcd, 0xfd, 0xd1, 0xee};
AES_KEY KLicenseeKey
int result = AES_set_decrypt_key(KLicenseeDecryptKey,0x80,&KLICENSEEKEY);
AES_decrypt(klicensee,&decryptedKLicensee,&KLicenseeKey);
EY is another key located inside the apploader and klicensee is the parameter.
Then we can finally remove the NPDRM layer using:
AES_KEY key; uint8_t iv[0x10]; memset(&iv[0],0,0x10); int result = AES_set_decrypt_key(&KLicenseeDecryptKey,0x80,&key); AES_cbc_encrypt(self + self->metaoffset + 0x20, self + self->metaoffset + 0x20,0x40,&key,&iv,0);
Observe the above code in action at [1]
Once that layer is removed we proceed as normal:
- Decrypt using AESCBC256 with the NPDRM keys to obtain the metadata keys
- Decrypt using AESCTR128 the data sha,hmac,iv keys
- Decrypt the data.
Source: http://www.ps3hax.net/showpost.php?p=259713&postcount=1 JuanNadie
Footnote: KLicenseeDecryptKey is located in appldr twice, e.g.
1.00:
Offset(h) 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
000187C0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
000187D0 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
000187E0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
000187F0 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
3.15:
Offset(h) 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
00018EB0 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
00018EC0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00018ED0 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
00018EE0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
3.55:
Offset(h) 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
00019730 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
00019740 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00019750 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
00019760 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
3.56:
Offset(h) 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
0001F920 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
0001F930 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
0001F940 F2 FB CA 7A 75 B0 4E DC 13 90 63 8C CD FD D1 EE òûÊzu°NÜ..cŒÍýÑî
0001F950 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
hex
appldr/npdrm 0.92-3.31: rev_0x01.rar (2.39 KB)
Part 2
Thank to all of you. I hope someone is able to code a program using this info.
First of all, I want to congratulate Euss of ps3devwiki on finding the klicensee decrypt key and provide a proof of concept of the AppLoder part of the algorithm. Check http://ps3devwiki.com/index.php?title=Talk:SELF_File_Format_and_Decryption Now you have the tools to decrypt all free executable content.
Euss they key is not duplicated... there are two cases that lead to the same (similar to the keys, two cases so two repeated tables).
Some of you asked what this algorithm is for. It has several use from backing up PSN games so they can be used with/without license (some countries allow backups, but NEVER sharing copyrighted material....) or use game updates on lower firmwares (some updates are NPDRM so they could not be decrypted and downgraded). I don't know if DUPLEX used this method or if they replaced the data with debug versions as some implied...
Also, it can be use is to modify geohot's make_self_npdrm to use non static keys for encoding. I don't know if that would be enough to make a self runnable on 3.56+ firmware. However it is a step on the right direction (I think extra modifications are required). If someone knows which parts of the self is whitelisted it would be an interesting addition to the thread. Sony was publishing 3.55 after 3.56 went online so I really interested to see which part of the SELF was whitelisted.
Others asked for the keys. I can not provide them nor functional code to avoid being sued... Graf and geohot were sued for providing the keys and/or functional code.
RIF key
However, I can provide a tip on getting the RIF key.... once decrypted bytes 0x40 to 0x4F should be xx xx xx xx xx xx xx xx xx xx xx xx 00 00 00 aa where x is random and aa is a number between 0x00 and 0x7F. It is located on the VSH.elf (remember that PPC64 has 8 byte aligment). That is a plaintext attack + dictionary(vsh). You don't need the curves as you can not sign rif nor act.dat (You can only check that file is valid). And the vsh keys can be easily find... graf chokolo called IDPS as device_id_ptr.... and the CONST is very near on code execution...
edit:
RIF's act.dat index decryption key
actdatix_dec_key : DA7D4B5E499A4F53B1C1A14A7484443B Actually in vsh.self
To yolbulduran: That is a piracy related question. In addition you have published confidential info, which anyone who does RCE should avoid (I do not have the SDK). The answer is NO. Why?. See this code:
ret = sceNpDrmVerifyUpgradeLicense("FAKE_CONTENT_TO_DETECT_CFW");
if(ret == 0){
banConsole();
print("CFW detected. Game will exit");
exit(-1);
}
ret = sceNpDrmVerifyUpgradeLicense("REAL_CONTENT");
if (ret == 0) {
int fd = openEncryptedContent("REAL_CONTENT.edat",......,keyForFile,sizeof(KeyForFile));
if (fd > 0) {
//Do things
}
}
First part is an example of how a developers can easily catch that modification and stop execution making it dangerous (could get a ban!!!). You modification says that the console has access to a fake content, which only CFW will have. When patching code the modification should be done only to the case you want to fix. That modification should go on the executable not on npd libraries. That way we do not patch the first verify but we will patch the second...
The second part is the real reason why it wont work... you REQUIRE the rif for opening the edat. The rif holds the klicensee for both SELF and EDAT. In fact I assume that the klicensee follows the same transformation upto the apploader. That key that you see on the command it is only used to check the HMAC on the NPD element (see geohot make_self_npdrm omac calculations)
For executable the problem is similar as when trying to run another PPU executable the program will finish and ask the vsh to run the other process which will undergo the full decryption algorithm... again you need the rif.
But.... what will happen if we decrypt the paid edat/SELF using the rif and then resign and encrypt as a free content before executing the code??? (Assuming we can sign edat)
WE CAN SIGN EXECUTABLES UPTO 3.55 THANKS TO FAIL0VERFLOW'S EPIC FAIL..... I think people do not really understands what that means...
Source: http://www.ps3hax.net/showpost.php?p=260574&postcount=8