subst reconstitution is broken

This commit is contained in:
Neale Pickett 2020-12-22 17:59:54 -07:00
commit 7390cb6f7c
4 changed files with 236 additions and 0 deletions

2
.gitignore vendored Normal file
View File

@ -0,0 +1,2 @@
decodes
data

10
Makefile Normal file
View File

@ -0,0 +1,10 @@
TARGETS += $(patsubst data/%.txt, decodes/%.csv, $(wildcard data/*.txt))
TARGETS += $(patsubst data/%.csv, decodes/%.csv, $(wildcard data/*.csv))
all: $(TARGETS)
decodes/%.csv: data/%.txt sunburst.py
./sunburst.py --outfile $@ $<
decodes/%.csv: data/%.csv sunburst.py
./sunburst.py --outfile $@ $<

24
NOTES.md Normal file
View File

@ -0,0 +1,24 @@
Stuff I still can't decode
====================
A lot of these things end in `.local`.
* fidelitycomm.lo|cal
* milledgeville.l|ocal
* cs.haystax.loc|al
* signaturebank.l|ocal
* vantagedatacente|rs.local
I bet these end in `.local` too:
* ABLE.loc|7l
* ETC1.loc|7l
* FVF.loca|m
* MOC.loca|m
* FSAR.LOC|7f
`7l = al`
`7f = AL`
`m = l`
How does `SCMRI.lo|ujjc` become `SCMRI.local`?

200
sunburst.py Executable file
View File

@ -0,0 +1,200 @@
#! /usr/bin/python3
# Neale Pickett <neale@lanl.gov>
# Unclassified/FOUO
#
# Created: 2020-12-14 16:49:51
# Last-modified: 2020-12-22 17:57:54
#
# Based on work by @RedDrip7 (twitter),
# who should be getting more credit in the English-speaking world.
import argparse
import base64
import codecs
import csv
import itertools
import re
import sys
knownDomains = [
"appsync-api.us-east-1.avsvmcloud.com",
"appsync-api.us-east-2.avsvmcloud.com",
"appsync-api.us-west-2.avsvmcloud.com",
"appsync-api.eu-west-1.avsvmcloud.com",
"avsvmcloud.com",
]
def xor(key, buf):
return bytes(b^k for b,k in zip(buf, itertools.cycle(key)))
Bsae32Alphabet = "ph2eifo3n5utg1j8d94qrvbmk0sal76c"
def DecodeBase32(s: str):
"""Not used by sunburst.
If Sunburst actually used Base32, this would work to decode things.
It doesn't work.
"""
t = s.translate(Bsae32Alphabet,
"ABCDEFGHIJKLMNOPQRSTUVWXYZ2345678")
while len(t) % 8 > 0:
t += '='
return base64.b32decode(t)
def DecodeBsae32(s: str):
"""Decode using zany base32-like algorithm.
The following opinion has been formed hastily and could be misinformed:
This is not proper Base32. It's more like somebody read about Base32,
implemented an encoder and decoder incorrectly, and went on to the next task.
"""
bits = 0
acc = 0
for c in s:
acc |= Bsae32Alphabet.find(c) << bits
bits += 5
out8 = []
while bits > 0:
out8.append(acc & 255)
acc >>= 8
bits -= 8
if bits:
del out8[-1]
return bytes(out8)
SubstitutionAlphabet = 'rq3gsalt6u1iyfzop572d49bnx8cvmkewhj'
SubstitutionXlat = str.maketrans(
SubstitutionAlphabet[4:] + SubstitutionAlphabet[:4], SubstitutionAlphabet)
SubstitutionXlat0 = str.maketrans(
SubstitutionAlphabet, ('0_-.' * 9)[:len(SubstitutionAlphabet)])
def DecodeSubst(s: str):
zeroBaby = False
out = []
for c in s:
if c == '0':
zeroBaby = True
continue
if zeroBaby:
out.append(c.translate(SubstitutionXlat0))
else:
out.append(c.translate(SubstitutionXlat))
zeroBaby = False
return ''.join(out)
PayloadsByGuid = {}
def DecodeDomain(domain: str):
s = domain.strip()
foundDomain = None
for d in knownDomains:
if s.endswith(d):
foundDomain = d
break
if not foundDomain:
raise RuntimeError("Can't find domain for %s" % s)
s = s[:-len(foundDomain)]
if not s:
return (None, None, "[no data transmitted]")
assert(s[-1] == '.')
s = s[:-1]
if len(s) < 16:
return (None, None, "[unable to decode: too short for any known decoding rules]")
eguid = DecodeBsae32(s[:15])
unknown_a = s[15]
guid = xor(eguid[0:1], eguid[1:])
s = s[16:]
payloads = PayloadsByGuid.setdefault(guid, [])
if s not in payloads:
if s.startswith("00"):
payloads.insert(0, s)
else:
payloads.append(s)
# People: friggin' preserve metadata, ugh.
# If I gave you every line of The Empire Strikes Back,
# sorted alphabetically, without timestamps,
# could you reconstruct the movie?
payloads.sort(key=len, reverse=True)
payload = ''.join(payloads)
if payload.startswith("00"):
buf = DecodeBsae32(payload[2:])
decoded = codecs.encode(buf, "quopri").decode("utf-8")
else:
decoded = DecodeSubst(payload)
return (guid, unknown_a, decoded)
class TextReader:
def __init__(self, infile):
self.infile = infile
self.fieldnames = ["name"]
def __iter__(self):
for s in self.infile:
yield {"name": s.strip()}
class CsvReader:
def __init__(self, infile):
self.reader = csv.DictReader(infile)
self.fieldnames = self.reader.fieldnames + ["guid", "unknown a", "decode"]
def __iter__(self):
for record in self.reader:
yield record
def main():
parser = argparse.ArgumentParser(
description="Decode sunburst Domain Generation Algorithm (DGA) names")
parser.add_argument("--text", dest="input", action="store_const", const="text",
help="Parse bambenek-style: list of fqdns, one per line")
parser.add_argument("--csv", dest="input", action="store_const", const="csv",
help="Parse CSV: records must be in a 'name' or 'fqdn' field")
parser.add_argument("infile", nargs="?",
type=argparse.FileType("r"), default=sys.stdin)
parser.add_argument("--outfile", nargs="?",
type=argparse.FileType("w"), default=sys.stdout,
help="CSV file to write (default stdout)")
args = parser.parse_args()
reader = None
if args.input == "text":
reader = TextReader(args.infile)
elif args.input == "csv":
reader = CsvReader(args.infile)
elif args.infile.name.endswith(".txt"):
reader = TextReader(args.infile)
elif args.infile.name.endswith(".csv"):
reader = CsvReader(args.infile)
else:
parser.print_help()
return
fieldnames = reader.fieldnames + ["guid", "unknown a", "decode"]
writer = csv.DictWriter(args.outfile, fieldnames)
writer.writeheader()
for record in reader:
name = record.get("name") or record.get("fqdn")
guid, unknown_a, ptext = DecodeDomain(name)
record["guid"] = int.from_bytes(guid or b"", "big")
record["unknown a"] = unknown_a
record["decode"] = ptext
writer.writerow(record)
if __name__ == '__main__':
main()