Skip to content

Instantly share code, notes, and snippets.

@williballenthin
Last active December 4, 2020 05:25
Show Gist options
  • Save williballenthin/3abc9577bede0aeef25526b201732246 to your computer and use it in GitHub Desktop.
Save williballenthin/3abc9577bede0aeef25526b201732246 to your computer and use it in GitHub Desktop.
generate a yara rule that matches the basic blocks of the current function in IDA Pro
'''
IDAPython script that generates a YARA rule to match against the
basic blocks of the current function. It masks out relocation bytes
and ignores jump instructions (given that we're already trying to
match compiler-specific bytes, this is of arguable benefit).
If python-yara is installed, the IDAPython script also validates that
the generated rule matches at least one segment in the current file.
author: Willi Ballenthin <william.ballenthin@fireeye.com>
'''
import logging
from collections import namedtuple
import idc
import idaapi
import idautils
logger = logging.getLogger(__name__)
BasicBlock = namedtuple('BasicBlock', ['va', 'size'])
# each rule must have at least this many non-masked bytes
MIN_BB_BYTE_COUNT = 4
def get_basic_blocks(fva):
'''
return sequence of `BasicBlock` instances for given function.
'''
ret = []
func = idaapi.get_func(fva)
if func is None:
return ret
for bb in idaapi.FlowChart(func):
ret.append(BasicBlock(va=bb.startEA,
size=bb.endEA - bb.startEA))
return ret
def get_function(va):
'''
return va for first instruction in function that contains given va.
'''
return idaapi.get_func(va).startEA
Rule = namedtuple('Rule', ['name', 'bytes', 'masked_bytes'])
def is_jump(va):
'''
return True if the instruction at the given address appears to be a jump.
'''
return idc.GetMnem(va).startswith('j')
def get_basic_block_rule(bb):
'''
create and format a YARA rule for a single basic block.
mask relocation bytes into unknown bytes (like '??').
do not include final instructions if they are jumps.
'''
# fetch the instruction start addresses
insns = []
va = bb.va
while va < bb.va + bb.size:
insns.append(va)
va = idc.NextHead(va)
# drop the last instruction if its a jump
if is_jump(insns[-1]):
insns = insns[:-1]
bytes = []
# `masked_bytes` is the list of formatted bytes,
# not yet join'd for performance.
masked_bytes = []
for va in insns:
size = idc.ItemSize(va)
if idaapi.contains_fixups(va, size):
# fetch the fixup locations within this one instruction.
fixups = []
fixupva = idaapi.get_next_fixup_ea(va)
fixups.append(fixupva)
# TODO: assume the fixup size is four bytes, probably bad.
fixupva += 4
while fixupva < va + size:
fixupva = idaapi.get_next_fixup_ea(fixupva)
fixups.append(fixupva)
# TODO: assume the fixup size is four bytes, probably bad.
fixupva += 4
# assume each fixup is four bytes (TODO!),
# and compute the addresses of each component byte.
fixup_byte_addrs = set([])
for fixup in fixups:
for i in range(fixup, fixup+4):
fixup_byte_addrs.add(i)
# fetch and format each byte of the instruction,
# possibly masking it into an unknown byte if its a fixup.
for i, byte in enumerate(idc.GetManyBytes(va, size)):
byte_addr = i + va
if byte_addr in fixup_byte_addrs:
bytes.append(ord(byte))
masked_bytes.append('??')
else:
bytes.append(ord(byte))
masked_bytes.append('%02X' % (ord(byte)))
elif 'call' in idc.GetMnem(va):
for i, byte in enumerate(idc.GetManyBytes(va, size)):
bytes.append(ord(byte))
masked_bytes.append('??')
else:
for byte in idc.GetManyBytes(va, size):
bytes.append(ord(byte))
masked_bytes.append('%02X' % (ord(byte)))
return Rule('$0x%x' % (bb.va), bytes, masked_bytes)
def format_rules(fva, rules):
'''
given the address of a function, and the byte signatures for basic blocks in
the function, format a complete YARA rule that matches all of the
basic block signatures.
'''
name = idc.GetFunctionName(fva)
# some characters aren't valid for YARA rule names
safe_name = name
BAD_CHARS = '@ /\\!@#$%^&*()[]{};:\'",./<>?'
for c in BAD_CHARS:
safe_name = safe_name.replace(c, '')
md5 = idautils.GetInputFileMD5()
ret = []
ret.append('rule a_{hash:s}_{name:s} {{'.format(
hash=md5,
name=safe_name))
ret.append(' meta:')
ret.append(' sample_md5 = "{md5:s}"'.format(md5=md5))
ret.append(' function_address = "0x{fva:x}"'.format(fva=fva))
ret.append(' function_name = "{name:s}"'.format(name=name))
ret.append(' strings:')
for rule in rules:
formatted_rule = ' '.join(rule.masked_bytes)
ret.append(' {name:s} = {{ {hex:s} }}'.format(
name=rule.name,
hex=formatted_rule))
ret.append(' condition:')
ret.append(' all of them')
ret.append('}')
return '\n'.join(ret)
def create_yara_rule_for_function(fva):
'''
given the address of a function, generate and format a complete YARA rule
that matches the basic blocks.
'''
rules = []
for bb in get_basic_blocks(fva):
rule = get_basic_block_rule(bb)
# ensure there at least MIN_BB_BYTE_COUNT
# non-masked bytes in the rule, or ignore it.
# this will reduce the incidence of many very small matches.
unmasked_count = len(filter(lambda b: b != '??', rule.masked_bytes))
if unmasked_count < MIN_BB_BYTE_COUNT:
continue
rules.append(rule)
return format_rules(fva, rules)
def get_segment_buffer(segstart):
'''
fetch the bytes of the section that starts at the given address.
if the entire section cannot be accessed, try smaller regions until it works.
'''
segend = idaapi.getseg(segstart).endEA
buf = None
segsize = segend - segstart
while buf is None:
buf = idc.GetManyBytes(segstart, segsize)
if buf is None:
segsize -= 0x1000
return buf
Segment = namedtuple('Segment', ['start', 'size', 'name', 'buf'])
def get_segments():
'''
fetch the segments in the current executable.
'''
for segstart in idautils.Segments():
segend = idaapi.getseg(segstart).endEA
segsize = segend - segstart
segname = str(idc.SegName(segstart)).rstrip('\x00')
segbuf = get_segment_buffer(segstart)
yield Segment(segstart, segend, segname, segbuf)
class TestDidntRunError(Exception):
pass
def test_yara_rule(rule):
'''
try to match the given rule against each segment in the current exectuable.
raise TestDidntRunError if its not possible to import the YARA library.
return True if there's at least one match, False otherwise.
'''
try:
import yara
except ImportError:
logger.warning("can't test rule: failed to import python-yara")
raise TestDidntRunError('python-yara not available')
r = yara.compile(source=rule)
for segment in get_segments():
matches = r.match(data=segment.buf)
if len(matches) > 0:
logger.info('generated rule matches section: {:s}'.format(segment.name))
return True
return False
def main():
va = idc.ScreenEA()
fva = get_function(va)
print('-' * 80)
rule = create_yara_rule_for_function(fva)
print(rule)
if test_yara_rule(rule):
print('success: validated the generated rule')
else:
print('error: failed to validate generated rule')
if __name__ == '__main__':
logging.basicConfig(level=logging.DEBUG)
logging.getLogger().setLevel(logging.DEBUG)
main()
@williballenthin
Copy link
Author

example rule:

rule D924C371DEC38EF931BE57FAE9B57AAE__LdrpCorInitialize0 {
  meta:
    sample_md5 = "D924C371DEC38EF931BE57FAE9B57AAE"
    function_address = "0x6a213856"
    function_name = "_LdrpCorInitialize@0"
  strings:
    $0x6a213856 = { 6A 20 68 ?? ?? ?? ?? E8 3E A0 06 00 B3 01 88 5D E7 68 ?? ?? ?? ?? E8 5F 56 01 00 33 F6 89 75 FC 8D 45 E0 50 56 56 6A 13 68 ?? ?? ?? ?? 56 E8 E7 53 01 00 BF 23 00 00 C0 3B C7 }
    $0x6a213896 = { C7 45 FC FE FF FF FF E8 D8 00 00 00 }
    $0x6a2138aa = { 68 30 00 FE 7F 8D 45 D8 50 E8 E8 DB 00 00 0F B7 7D D8 83 C7 2C 81 FF FE FF 00 00 }
    $0x6a2138cb = { A1 ?? ?? ?? ?? 05 00 00 18 00 57 50 FF 35 ?? ?? ?? ?? E8 3E B2 01 00 89 45 D4 85 C0 }
    $0x6a2138ed = { 33 C0 66 89 45 D0 66 89 7D D2 8D 45 D8 50 8D 45 D0 50 E8 8C E8 01 00 68 ?? ?? ?? ?? 8D 45 D0 50 E8 7E E8 01 00 68 ?? ?? ?? ?? 8D 45 D0 50 E8 70 E8 01 00 8D 45 E0 50 8D 45 D0 50 56 56 E8 31 85 02 00 8B F8 FF 75 D4 56 FF 35 ?? ?? ?? ?? E8 20 96 01 00 85 FF }
    $0x6a213944 = { 8D 45 DC 50 56 68 ?? ?? ?? ?? FF 75 E0 E8 4A 00 00 00 8B F0 85 F6 }
    $0x6a21395c = { 8B 15 30 03 FE 7F 8B CA 83 E1 1F 33 55 DC D3 CA 89 15 ?? ?? ?? ?? }
    $0x6a213974 = { E8 6C 9F 06 00 C3 }
    $0x6a21397a = { 68 ?? ?? ?? ?? E8 1C 58 01 00 C3 }
    $0x6a213989 = { FF 75 E0 E8 0F 38 02 00 }
    $0x6a287c44 = { 8D 45 E0 50 56 56 6A 0F 68 ?? ?? ?? ?? 56 E8 19 10 FA FF 3B C7 0F 94 C0 FE C8 22 D8 88 5D E7 }
    $0x6a287c68 = { 33 F6 8A 5D E7 }
    $0x6a287c72 = { B8 06 01 00 C0 }
    $0x6a287c7c = { B8 17 00 00 C0 }
    $0x6a287c86 = { 8D 45 E0 50 68 ?? ?? ?? ?? 56 56 E8 CA 41 FB FF 85 C0 }
  condition:
    all of them
}

@williballenthin
Copy link
Author

future development tracked here: https://github.com/williballenthin/idawilli

@mrexodia
Copy link

mrexodia commented Sep 2, 2017

I ported this script to an x64dbg plugin: https://github.com/mrexodia/YaraGen

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment