Subject: The "Twelve Tricks" trojan - alert and description Date: 12 Feb 90 21:31:29 GMT Sender: daemon@ucbvax.BERKELEY.EDU The "Twelve Tricks" trojan - alert and description We have recently received and analysed a trojan that we believe warrants an urgent alert. We are calling it the Twelve Tricks trojan, and it is very interesting, very nasty, and quite complex. This message is not meant to be a complete description of the trojan - we feel that it is important to get a warning out quickly, rather than aim for completeness. It is not a virus. The trojan consists of a program (more about this aspect later) which you run; running the program, as well as the obvious things that the program is expected to do, also replaces the partition record (also called the Master Boot Record, or MBR) on your hard disk with its own version. This can easily be recognised by inspecting the hard disk at cylinder zero, head zero, sector one, which can be done with a disk sector editor such as Peeka. If the partition has this trojan in place, it will contain the following text near the beginning: SOFTLoK+ V3.0 SOFTGUARD SYSTEMS INC 2840 St. Thomas Expwy,suite 201 Santa Clara,CA 95051 (408)970-9420 At this point, let us state that we believe that the company mentioned above has nothing whatsoever to do with the trojan; perhaps the trojan author has a grudge against them. The trojan uses a far call to the hard disk Bios code in order to plant this partition. To do this, it must know the location in memory of the entry point; it tries five different ones, one of which is the one documented in the IBM PC-XT Technical reference manual, and the other four are presumably fairly common alternatives. The purpose of planting the trojan with a far call is, we believe, to escape detection by Active Monitor programs that protect a computer by monitoring the interrupt table, and preventing unauthorised writes to system areas on the hard disk. Since Twelve Tricks doesn't use an interrupt to plant the MBR, such programs won't be able to prevent it. We tested this using Flushot+, probably the most successful of the Active Monitors, and Twelve Tricks went straight through it - the same would be true, we think, of any other Active Monitor. The Replacement MBR When the MBR is run, which is every time you boot from the hard disk, Twelve Tricks copies 205 (d7h) bytes of itself onto locations 0:300h to 0:3d6h. This overwrites part of the interrupt vector table, but it is a part that doesn't get used very much. This means that these d7h bytes are memory resident without having to use any of the TSR calls of Dos, and without having to reserve part of high memory. Reserving part of high memory is the usual ploy used by Boot Sector Viruses, but the drawback of that route is that you might notice that a few kb from your 640 kb has disappeared (CHKDSK would reveal this). The method used by Twelve Tricks would not show up as a loss from your 640 kb. When the computer is started up, a random number generator determines which of the Twelve Tricks will be installed. It does the installation by replacing one of the interrupt vectors with a vector that points to the Twelve Tricks own code, and then chains on to the original code. The twelve tricks are: 1. Insert a random delay loop in the timer tick, so that 18.2 times per second, the computer executes a loop that is randomly between 1 and 65536 long (different each time it is executed). This slows the machine down, and makes it work rather jerkily. 2. Insert an End-Of-Interrupt in the timer tick. This interferes with the servicing of hardware interrupts, so for example, the clock is stopped, TSRs that depend on the timer tick don't work, and the floppy motor is permanently on. 3. Every time a key is pressed or released, the timer tick count is incremented by a random number between 0 and 65535. This has a variety of effects; programs sometimes won't run, when you type "TIME" you get "Current time is divide overflow", and copying files sometimes doesn't work. 4. Every time interrupt 0dh is executed, only do the routine three times out of four. Interrupt 0dh is used on PCs and XTs for the fixed disk, on ATs for the parallel port. 5. Every time interrupt 0eh is executed, only do the routine three times out of four. Interrupt 0eh is used for the floppy disk. 6. Every time interrupt 10h is called (this is the video routine), insert a delay loop that is randomly between 1 and 65536 long (different each time it is executed). This slows the video down, and makes it work rather jerkily and/or slowly. 7. Every time the video routine to scroll up is called, instead of the requested number of lines being scrolled, the entire scrolling window is blanked. 8. Every time a request is made to the diskette handler, it is converted into a write request. This means that the first time you try to read or write to a diskette, whatever happens to be in the buffer will be written to the diskette, and will probably overwrite the boot sector, FAT or directory, as these must be read before anything else can be done. If you try to read a write protected diskette, you get "Write protect error reading drive A". If you do a DIR of a write enabled diskette, you get "General Failure ...", and if you inspect the diskette using a sector editor, you'll find that the boot and FAT have been zeroed or over-written. 9. Every time interrupt 16h is called (read the keyboard) the keyboard flags (Caps lock, Num lock, shift states etc) are set randomly before the keystroke is returned. This means that at the Dos prompt, the keyboard will only work occasionally. Programs that poll interrupt 16h will be unusable. Holding down the Del key will trigger a Ctrl-Alt-Del. 10. Everything that goes to the printer is garbled by xoring it with a byte from the timer tick count. 11. Every letter that is sent to the printer has its case reversed by xoring it with 20h. Also, non-alpha characters are xored, so a space becomes a null, and line feeds don't feed lines. 12. Whenever the Time-Of-Day interrupt (1ah) is executed, do an End-Of-Interrupt instead. This means that you can't set the system clock, and the time is set permanently to one value. These are the twelve tricks. In addition there are two more things that the trojan does. It uses a random number generator; one time out of 4096, it does a low level format of the track that contains the active boot sector; this will also destroy part of the first copy of the FAT. You can recover from this by creating a new boot sector, and copying the second copy of the FAT back over the first copy. After it does the format, it will display the message "SOFTLoK+ " etc as above, and hang the computer. If it doesn't do the format, it makes a random change to a random word in one of the first 16 sectors of the FAT, which will make a slight and increasing corruption in the file system. This is perhaps the worst of the things that it does, as it will cause an increasing corruption of the files on the disk. The Dropper program The program that drops the trojan was, in the specimen that we analysed, a hacked version of CORETEST, a program to benchmark hard disk performance. The file is CORETEST.COM, it is version 2.6, (dated 1986 in the copyright message) had a length of 32469 bytes, and it was timestamped 6-6-86, 9:44. When we looked in more detail at this program, we found some interesting things. It looks as if the original CORETEST program was an EXE file, and the trojan author prepended his code to it. This code consists of some relocation stuff, then a decryptor, to decrypt the following 246h bytes. The decryption is a double xor with a changing byte. Those 246h bytes, when run, examine the memory to try to find one of five sets of hard disk handler code (presumably corresponding to five Bioses). When it finds one of them, (we have identified the first one as being the IBM XT Bios) it plants the trojan MBR in place, using a far call to the Bios code. The trojan MBR is 200h of the 246h bytes. The trojan is patched so that it also does disk accesses using a far call to the same location. Finally, the prepended trojan passes control to the original program. We call the combination of the prepended code, plus the original program, the Dropper. The main purpose of the encryption, we would guess, is to evade detection by programs that check code for bombs and trojans. There are no suspicious strings or interrupt calls in the code until it is decrypted at run time. As far as we can tell, it is not a virus, but a trojan. However, it is unlikely that all the patching to the original program was done by hand - it is far more likely that the trojan author wrote a prepender program (we would call this the Prepender), to automatically attach his code to the target executable. If this is the case, then there are two consequences. The first is that he might have trojanised other programs besides the one that we have examined. In other words, there might be other Droppers around besides the one we have examined. The second is that if that is the case, we cannot rely on the encryption having the same seed each time, as the Prepender might change the seed each time it operates. So it would be unsafe to search files for the encrypted MBR. Instead, we propose a search string based on the decryptor. Indeed, a further possibility exists. The Prepender program might have been placed into circulation, and people running it would unwittingly be creating additional Droppers. There is absolutely no evidence to suggest that that is actually the case, but we would ask anyone who detects this Dropper in one of their files, to also examine all the others. Detection Here's a variety of ways to detect the trojan. The hexadecimal string e4 61 8a e0 0c 80 e6 61 is to be found in the MBR. This string will also be found in memory if you have booted from a trojanised MBR, at location 0:38b. You can use Debug to search in memory. A useful search string to detect the Dropper is be 64 02 31 94 42 01 d1 c2 4e 79 f7 Getting rid of it It's easy to get rid of Droppers; just delete them and replace them with a clean copy. If you find the string above in the MBR or in memory at 0:38b, you need to boot from a clean Dos diskette and replace the partition record. DO NOT use Fdisk to do this unless you are prepared for Fdisk to zero your FAT and directory; you will lose all your data that way. One way would be to do a file-by-file backup, low-level format to get rid of the trojan MBR, then Fdisk Format and restoer your backup. We would recommend doing two backups using as different methods as possible if you use this route, in case one of them fails to restore. The other way to replace the partition is to run a program that drops a clean partition record onto the MBR, but doesn't change the partitioning data. We are currently preparing one of these - please ask if you need it. Damage done The whole of the MBR is used for the code. Most normal MBRs don't use more than half the space, and a number of other programs have started using this space. For example Disk Manager, and the Western Digital WDXT-Gen controllers (but the Dropper doesn't work on the WDXT-Gen). This means that the Dropper might cause an immediate problem in some circumstances. The main damage done, however, will be in the impression that this trojan creates that your hardware is suffering from a variety of faults, which usually go away when you reboot (only to be replaced by other faults). Also, the FAT gets progressively corrupted. Occurrences So far, this has only been reported in Surrey, England. It was noticed because it made a disk using Speedstor to control it, non-bootable. Disks that are controlled in the normal way, remain bootable. We would be grateful if any sightings could be reported to us, especially if the Dropper program is different from the one we have examined; we would also like a specimen of it, Please report instances to the addresses below: Dr Alan Solomon Day voice: +44 494 791900 S&S Anti Virus Group Eve voice: +44 494 724201 Water Meadow Fax: +44 494 791602 Germain Street, BBS: +44 494 724946 Chesham, Fido node: 254/29 Bucks, HP5 1LP Usenet: drsolly at ibmpcug.co.uk England Gold: 83:JNL246 CIX, CONNECT drsolly or Mr Christoph Fischer Day voice: +49 721 6084041 Micro-BIT Virus Centre Eve voice: +49 721 861540 University of Karlsruhe Fax: +49 721 621479 Zirkel 2 BITNET: RY15 at DKAUNI11 D-7500 Karlsruhe 1 West-Germany