I've written a backup script to make a backup of the latest modified files in 12 hours. Basically, it searches two directories for modified files (statically coded) then the find command is used to find the related files. A log file is then created to keep a list of the files. Later on, they are used to form a list of files and these files are used for the backup by tar.

Let me know which parts could be developed so that I can improve on and learn new things.

#!/usr/bin/env python
''' backup script to make a backup of the files
    that are newer than 12 hours in the 
    home directory 
    " man find "
    " man tar  "
    for more information on the dates and the specific options
    on how to set different specifications for the selections
import sys
import os
import time

if (len(sys.argv) != 2):
  # if argument count is different 
  # than 2, then there is an error  
  print "Error on usage -> binary <hours_count_back>"
  # convert to integer value
  hours_back = int(sys.argv[1])

message_start = "Operation is starting".center(50, '-')
message_end   = "Operation terminated with success".center(50,'-')
print message_start
# source directories to check for backup
source = [ '/home/utab/thesis', '/home/utab/Documents' ]
# find files newer than 24 hours
# change to the target directories
newer_files = []
log_file = "log.dat"
# cmd to find newer files
# than 24 hours
cmd_find = "find . -type f -a -mmin " + str(-hours_back*60) + " > " + log_file
for directory in source:
  # iterate over the directories
  # change to the directory first
  # apply the command
  # files are found with respect to the current directory
  # change the "." to directory for correct backups
  c = 0
  # process log file
  files = []
  lines = []
  log_in = open(log_file,'r')
  # read lines without \n character
  while 1:
      l = log_in.readline()
      if not l: 
      # do not include the newline
      # -1 is for that
  for l in lines:
      l_dummy  = l.replace( '.', directory, 1 )
  # extend the list with newer files
#print newer_files
print newer_files
# date
today =  time.strftime('%Y%m%d')
# current time of the date
# possible to do different backups in different times of the day
now   = time.strftime('%H%M%S')
target_directory = "/home/utab/"
target = target_directory + today + "_" + now + \

# do the actual back up
backup_cmd = "tar -C ~ -zcvf %s %s" % ( target , ' '.join(newer_files) )
status = os.system(backup_cmd)

if status == 0:
    print message_end
    print "Back-up failed"

2 Answers 2


The script is vulnerable to an injection attack. If anyone can write files under any of the directories to be backed up then the script can be made to execute code of their choosing. (I realize that this is your home directory and "this could never happen" but bear with me for a moment). The problem is that a non-validated string is passed to os.system():

backup_cmd = "tar -C ~ -zcvf %s %s" % ( target , ' '.join(newer_files) )
status = os.system(backup_cmd)

Attacker just needs to:

cd some/where/writable
touch '; rm -rf ~'

But it doesn't even need to be a malicious attack. If any of the files have characters in their names that are interesting to the shell then the script may fail. A tamer example: if any filename contains a hash character (#) then all subsequent filenames in the newer_files list will be ignored and will not be backed up by tar(1).

Instead please consider using os.exec*() or better still the subprocess module. These take a vector of arguments rather than a string and so do not suffer from the same problem.


A general suggestion which might yield a performance boost: if you can somehow determine that the directories in the source lists are on different harddisks, you could execute the main loop (which spends most of the time in the find invocation, I guess) in parallel.

That being said, you could replace

for l in lines:
    l_dummy  = l.replace( '.', directory, 1 )
# extend the list with newer files

with a list comprehension:

newer_files.extend([l.replace( '.', directory, 1 ) for l in lines])

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.