sqlmap/ok.py

143 lines
5.0 KiB
Python
Raw Normal View History

import subprocess
from datetime import datetime
import sys
import os
import csv
import re
from urllib.parse import urlparse
#****************************************************************************************
# this function verify if the argument given is an URL
def est_url(url):
try:
result=urlparse(url)
return all([result.scheme, result.netloc])
except Exception as e:
return False
# this function verify if the argument given is a request
def est_requete(req):
exp_reg= expression_reguliere = re.compile(
r'\b(SELECT|UPDATE|DELETE|INSERT INTO|CREATE|ALTER|DROP|GRANT|REVOKE)\b',
re.IGNORECASE)
return re.search(exp_reg,req) is not None
#*****************************************************************************************
# read the arguements
arg=sys.argv
#*****************************************************************************************
#verify if the arguments given are correct or not
if (('-u' in arg)and('--sql-query' in arg)):
if (est_url(arg[2])):
if (est_requete(arg[-1])):
commande_sqlmap = ['python','sqlmap.py']+arg[1:]
else:
print("verify your request")
exit()
else:
print("verify your URL")
exit()
else:
print("Verify that you input correctly -u or --sql-query")
exit()
#*****************************************************************************************
#*****************************************************************************************
#extract the name of columns from the request
deb="SELECT"
fin="FROM"
#find the start point of the ""deb"" variable in the request
temp1=(arg[-1].upper()).find(deb)
#find the start point of the ""fin"" variable in the request
temp2=(arg[-1].upper()).find(fin)
#select the request from the arguments given bu the user in command line
temp_res=arg[-1]
#start selection of the columns names after the ""deb"" variable
T=len(deb)-temp1+1
temp_res=temp_res[T:temp2]
#split the String that contains columns name
temp_res=(temp_res.strip()).split(',')
#******************************************************************************************
#******************************************************************************************
#process the request
res=subprocess.check_output(commande_sqlmap)
#res=subprocess.run(commande_sqlmap,capture_output=True, text=True)
#view the result of processing the resquest
res = res.decode('utf-8')
print(res)
# Select actuel time
heure_actuelle = datetime.now()
heure_formattee = heure_actuelle.strftime("%H:%M:%S")
# ""m_deb"" variable is a string that represents the start point of showing the result of the request
m_deb="["+str(heure_formattee)+"]"+" [INFO] fetching SQL SELECT statement query output:"
# ""m_fin"" variable is a string that represents the end point point of showing the result of the request
m_fin="["+str(heure_formattee)+"]"+" [INFO] fetched data logged to text files under"
#""debut_message"" contains the position of the ""m_deb"" variable in the result
debut_message = res.find(m_deb)
#""fin_message"" contains the position of the ""m_fin"" variable in the result
fin_message = res.find(m_fin)
#select a part of the result that is between ""m_deb"" variable and ""m_fin"" variable
message = res[debut_message:fin_message].strip()
lignes = message.splitlines()
# ""donnees_sauvegarder"" variable is a list of dictionaries
donnees_sauvegarder = []
# loop on the ""lignes"" variable
for ligne in lignes:
# verify if this line contain [*]
if ligne.startswith("[*]"):
# extract data after '[*]'
v_temp=(ligne[4:].strip()).split(',')
d_temp={}
for i in range(len(temp_res)):
# add for every column from ""temp_res"" variable a value from ""v_temp"" variable
d_temp.update({temp_res[i]:v_temp[i]})
# add the dictionarie in the list
donnees_sauvegarder.append(d_temp)
#*****************************************************************************************************
# create a file named ""data_save.csv"" in same repositories with this script file
file_csv="data_save.csv"
if(len(donnees_sauvegarder)!=0):# if the process of the request send data
with open(file_csv,mode='w',newline='') as file_csv:
# write the header of this CSV file
writer=csv.DictWriter(file_csv,fieldnames=temp_res)
writer.writeheader()
# loop on the list of dictionaries named ""donnees_sauvegarder"" to wite its content on the CSV file
for part in donnees_sauvegarder:
writer.writerow(part)
# if the data is successfully saved in the CSV file this message show the place of the file
print("data saved in csv format under "+"'"+os.path.realpath("data_save.csv")+"'")
else:
print("data don't exist to create CSV file")
#********************************************************************************************************