Last active
November 21, 2022 13:37
-
-
Save DarkCoderSc/de3b2a320fb3c72037ed0eaecac239e6 to your computer and use it in GitHub Desktop.
Extract ASCII and/or UNICODE strings from files.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python3 | |
# Jean-Pierre LESUEUR (@DarkCoderSc) | |
# https://keybase.io/phrozen | |
import argparse | |
import mmap | |
from itertools import chain | |
def extract_strings(file, min_length=4, unicode=False): | |
printable_ascii = b"0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~ \t" | |
if unicode: | |
char_size = 2 | |
else: | |
char_size = 1 | |
with mmap.mmap(file.fileno(), length=0, access=mmap.ACCESS_READ) as mmap_obj: | |
string = b"" | |
offset = 0 | |
for cursor in range(0, mmap_obj.size(), char_size): | |
b = mmap_obj.read(char_size) | |
if b[0] in printable_ascii: | |
if char_size == 2 and b[1] != 0: | |
continue | |
string += b[0].to_bytes(1, byteorder='big') | |
else: | |
if len(string) >= min_length: | |
yield offset, string.decode('ascii').replace("\t", "") | |
string = b"" | |
offset = cursor | |
if __name__ == "__main__": | |
parser = argparse.ArgumentParser(description=f"Binary String Extractor") | |
parser.add_argument('-f', '--file', type=argparse.FileType('rb'), dest="file", required=True, help="Binary file to inspect for strings.") | |
parser.add_argument('-o', '--offset', default=False, dest="show_offset", action="store_true", help="Show string location in file (string offset).") | |
parser.add_argument('-l', '--min-length', default=4, required=False, dest="min_length", action="store", help="Minimum length of extracted string.") | |
parser.add_argument('-m', '--extract-mode', dest="mode", default='all', choices=['all', 'ascii', 'unicode'], help="Filter string extraction by its encoding nature.") | |
try: | |
argv = parser.parse_args() | |
except IOError as e: | |
parser.error() | |
ascii_strings = iter([]) | |
unicode_strings = iter([]) | |
if argv.mode == "all" or argv.mode == "ascii": | |
ascii_strings = extract_strings(argv.file, argv.min_length) | |
if argv.mode == "all" or argv.mode == "unicode": | |
unicode_strings = extract_strings(argv.file, argv.min_length, True) | |
for offset, string in chain(ascii_strings, unicode_strings): | |
if argv.show_offset: | |
print("{} : {}".format( | |
offset, | |
string, | |
)) | |
else: | |
print(string) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment