Reversing Thief and the current Eidos protection scheme
by Kilby
9 January 1999


academy
Back to the academy

I would like to say a few words at the start of this file as I feel that it is necessary to do so

Over the last year or so I have been looking at Fravia+'s site, and thinking that it was good thing that somebody has taken the effort to tutor the new generation.

I started taking other peoples code apart in 1983, yet I have been retired for around five/six years... Why did I decide to return ?

Well I think this will be a bit of a one off to be honest.

There seems to be a shortage of good information on reversing CD based protection systems, I don't know if this counts as good information but it may help somebody.

After checking out Fravia+'s site every once in a while I thought I should add a tutorial, to give something back for the pleasure I have had from reading some of the essays.

During the summer I finally moved on from my old 486 to a decent spec system, and also got myself a CD-R.

Well after playing about with it for a while, I started to look into what the best mastering software was, and found the CD-R discussion forum at www.cdrom-guide.com, where everybody had been complaining about how they couldn't copy certain games.

Well amongst the usual no brain postings there where a few interesting reports, such as about CDs over 75 minutes, illegal Table of Contents and large files on disk.

Whenever you saw a patch listed it seemed that the inhabitants of the message board, thought of the author as some sort of supernatural creature.

The reality is that our brains are wired a little differently, to everybody else.

Most of the protection schemes have the following simple structure:
Is the drive a CD ?
Does a file exist ?
Can that drive be written to ?

That's about it, a bit lame really, perhaps even hiding this inside a DLL, if they are in an ugly mood.

Well considering I wanted Thief (I collect as much as I can afford by ex Infocom staff), and Eidos used what looked like a "mean" protection system. I bought Thief.

Well... it was quite a disappointment to tell the truth (not the game, I may add it puts a whole new spin on the Doom/Quake genre, but the protection itself).

The protection system used for Thief relies upon :

1: More than 75 mins of data on the CD
2: An invalid TOC, Data, Audio, Audio, Data
3: Four Fake files on the first data track, each of approx. 515 MB, KXEP.AFP, AZXP.AFP, TFJB.AFP & IGYQ.AFP.
4: Mixed case in the CD Volume name.

Anyway the Eidos TOC is only to make a straight copy difficult, as the two audio tracks aren't used at all and the other tracks isn't even checked for.

This is real dumb as all the necessary functions are there in MSCDEX.

Without these three tracks, 74 minutes is more than enough to hold the real data track.

The four fake .AFP files are not used properly within the protection system, although they deter casual copying, there is a 4KB block loaded, and I think they even check it's the correct data, it is not necessary for the program to actually run.

Obviously it would have been easy enough to copy each of the 4096 byte blocks to separate files and simply seek to offset 0 (rather than 696406016, 496444928, 696514560 or 696346624, as actually used), and then read the 4096 bytes each time, but it would have slowed me down.

To be honest it did slow me down because I was convinced that they would not be so stupid and lazy not to use this data (I was wrong).

I don't even think they check for the mixed case in the CD volume label.

Tools used:
Filemon, W32Dasm and hiew

Using filemon, I watched as thief.exe opened each of these hidden .AFP files where opened and a seek was carried out to around 696496016 bytes into the file, 4096 bytes read and then the file closed.

This occurred for each of the .AFP files, with various (but similar seek values).

After that I cut a CD with all the files except the .AFP fakes, and as the software I was using wouldn't allow me to use mixed case volume labels

I used W32Dasm to disassemble thief.exe.

I searched for .AFP and found that one large subroutine was used to open them, this made life easier.

The routine was at 00526100 through to 526274 and was called from 3 addresses, all within a 4KB region of RAM.

Then I searched for GetDriveTypeA which lives between 0050d720 to 0050d735

It called from 0050d773 and 0050da5e

The call to 0050d773 is placed within a routine (between 0050d740 to 0050d78f)

Appeared to be reserved for scanning through drives D: to Z; looking for CD drives.

Not very interesting

The call from 0050da5e looked much more interesting as it seemed to test the return value of GetDriveTypeA.

Six instructions after this call to GetDriveTypeA was a call to 00526100 the .AFP read routine.

A few instructions further on a second call to the .AFP routine took place and some testing of eax the routine did a ret

This routine is called from 0050dadc

The other call to the .AFP read appears to trace back to 0050dad5

This was the region that the tests on the invisible .AFP files are called from

This means that the all AFP reads and protection checks take place from an area separated by 2 bytes.

Just like this
:0050DAD5 E8E6FEFFFF  call 0050D9C0 ; First Call
:0050DADA EB05   jmp 0050DAE1
:0050DADC E82FFFFFFF  call 0050DA10; Second and third calls
To quote Homer Simpson DOH!

The Subroutine this is in looks just like this
* Referenced by a CALL at Address:00414FCC

:0050DAC0 6A00   push 00000000
:0050DAC2 6A00   push 00000000

* Possible StringData Ref from Data Obj ->"only_check_path"

:0050DAC4 68E0BF6000  	push 0060BFE0
:0050DAC9 E862AF0300  	call 00548A30
:0050DACE 83C40C   	add esp, 0000000C
:0050DAD1 84C0   	test al, al
:0050DAD3 7507   	jne 0050DADC
:0050DAD5 E8E6FEFFFF  	call 0050D9C0
:0050DADA EB05   	jmp 0050DAE1

* Referenced by a Jump at Address:0050DAD3(C)

:0050DADC E82FFFFFFF  	call 0050DA10

* Referenced by a Jump at Address: 0050DADA(U)

:0050DAE1 85C0   	test eax, eax
:0050DAE3 7405   	je 0050DAEA
:0050DAE5 E986FBFFFF  	jmp 0050D670

* Referenced by a Jump at Address: 0050DAE3(C)

:0050DAEA 33C0   	xor eax, eax
:0050DAEC C3   		ret
Well doesn't :00414FCC look a bit out of place in all this, and just look at this
:0050DAC0 6A00   push 00000000
:0050DAC2 6A00   push 00000000
Looks like some .OBJ code has been linked, and they wanted to make sure of some free space on the stack.

Well if we look at 00414FCC, we could nop the call and see what happens.

But I wont.

Why ?

Well assuming that the protection code called is not written by Looking Glass (the authors of Thief), as all the Eidos stuff on release, has similar (if not the same) protection.

They won't check the internal code of the protection routines.

However they may check their own code, including the call to the disk protection.

So rather than using four NOPs to delete the call 0050DAC0, I chose to add a RET (C3) at 0050DAC0, that way all the code by Looking Glass is totally intact.

So fire up a hex editor and search for
6A,00,6A,00,68,E0,BF,60,00
Change the (first) 6A to C3, it is the only occurrence of these hex values, and write the data back to disk.

The file offset is 10CEC0, (if you are using hiew then it is 50DAC0, in the other offset mode)

Feel free to skip the history section that follows.

This is a bit of history:
I have been retired from cracking for around five/six years.
I started cracking in 1983
I worked my through a range of machines, from the VIC 20, Spectrum, Amstrad CPC, Atari ST, Amiga & PC, which I owned.
I also cracked on machines that I didn't own for the challenge, such as the C64, Oric, and BBC (Acorn), just for the challenge.
When I wasn't doing that, photocopiers, locks, phones, conjuring and juggling and anything I wasn't supposed to be able to do became targets.
About five years ago I got bored, Windoze had turned up, games where boring and everything was being written in C, and DOS extenders just took the fun out of it all, it was just too easy to hide a small routine within the morass that C compilers created.
Real life also caught up so since then I have been playing with Ultra Sparcs, and DEC Alphas, and trying to get enough time to sleep.
This program took far to long to reverse because of never reversing a windows program, being totally out of practice, overestimating the enthusiasm of the protection authors and finally but not least real life.

Anyway I have now spilled a beer over some listings I have beside me, so it's time to send this small essay to fravia+ and fuck off to bed.

Regards to all

Kilby...
You'r deep inside fravia's pages of reverse engineering, choose your way out
redhomepage red links red anonymity red+ORC redstudents' essays redtools redcocktails
redacademy database redantismut search_forms redmail_fravia
redis reverse engineering legal?