mirror of
https://github.com/The-Art-of-Hacking/h4cker
synced 2024-11-24 11:53:02 +00:00
366 lines
16 KiB
Python
366 lines
16 KiB
Python
#!/usr/bin/env python
|
|
#------------------------------------------------------------------
|
|
# November 2014, created within ASIG
|
|
# Author James Spadaro (jaspadar)
|
|
# Co-Author Lilith Wyatt (liwyatt)
|
|
#------------------------------------------------------------------
|
|
# Copyright (c) 2014-2017 by Cisco Systems, Inc.
|
|
# All rights reserved.
|
|
#
|
|
# Redistribution and use in source and binary forms, with or without
|
|
# modification, are permitted provided that the following conditions are met:
|
|
# 1. Redistributions of source code must retain the above copyright
|
|
# notice, this list of conditions and the following disclaimer.
|
|
# 2. Redistributions in binary form must reproduce the above copyright
|
|
# notice, this list of conditions and the following disclaimer in the
|
|
# documentation and/or other materials provided with the distribution.
|
|
# 3. Neither the name of the Cisco Systems, Inc. nor the
|
|
# names of its contributors may be used to endorse or promote products
|
|
# derived from this software without specific prior written permission.
|
|
#
|
|
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS "AS IS" AND ANY
|
|
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
|
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
|
# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY
|
|
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
|
|
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
|
|
# ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
|
|
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
#------------------------------------------------------------------
|
|
# Type definitions for the fuzzer
|
|
#
|
|
# This script defines the various message and data types used in
|
|
# the fuzzer, and utility functions used by them.
|
|
#------------------------------------------------------------------
|
|
|
|
class MessageSubComponent(object):
|
|
def __init__(self, message, isFuzzed):
|
|
self.message = message
|
|
self.isFuzzed = isFuzzed
|
|
# This includes both fuzzed messages and messages the user
|
|
# has altered with messageprocessor callbacks
|
|
self._altered = message
|
|
|
|
def setAlteredByteArray(self, byteArray):
|
|
self._altered = byteArray
|
|
|
|
def getAlteredByteArray(self):
|
|
return self._altered
|
|
|
|
def getOriginalByteArray(self):
|
|
return self.message
|
|
|
|
# Contains all data of a given packet of the session
|
|
class Message(object):
|
|
class Direction:
|
|
Outbound = "outbound"
|
|
Inbound = "inbound"
|
|
|
|
class Format:
|
|
CommaSeparatedHex = 0 # 00,01,02,20,2a,30,31
|
|
Ascii = 1 # asdf\x00\x01\x02
|
|
Raw = 2 # a raw byte array from a pcap
|
|
|
|
def __init__(self):
|
|
self.direction = -1
|
|
# Whether any subcomponent is fuzzed - might not be entire message
|
|
# Default to False, set to True as message subcomponents are set below
|
|
self.isFuzzed = False
|
|
# This will be populated with message subcomponents
|
|
# IE, specified as message 0 11,22,33
|
|
# 44,55,66
|
|
# Then 11,22,33 will be subcomponent 0, 44,55,66 will be subcomponent 1
|
|
# If it's a traditional message, it will only have one element (entire message)
|
|
self.subcomponents = []
|
|
|
|
def getOriginalSubcomponents(self):
|
|
return [subcomponent.message for subcomponent in self.subcomponents]
|
|
|
|
# May or may not have actually been changed
|
|
# Version of subcomponents that includes fuzzing and messageprocessor changes from user
|
|
# Is transient and reverted to original every iteration
|
|
def getAlteredSubcomponents(self):
|
|
return [subcomponent.getAlteredByteArray() for subcomponent in self.subcomponents]
|
|
|
|
def getOriginalMessage(self):
|
|
return bytearray().join([subcomponent.message for subcomponent in self.subcomponents])
|
|
|
|
# May or may not have actually been changed
|
|
# Version of message that includes fuzzing and messageprocessor changes from user
|
|
# Is transient and reverted to original every iteration
|
|
def getAlteredMessage(self):
|
|
return bytearray().join([subcomponent.getAlteredByteArray() for subcomponent in self.subcomponents])
|
|
|
|
def resetAlteredMessage(self):
|
|
for subcomponent in self.subcomponents:
|
|
subcomponent.setAlteredByteArray(subcomponent.message)
|
|
|
|
# Set the message on the Message
|
|
# sourceType - Format.CommaSeparatedHex, Ascii, or Raw
|
|
# message - Message in above format
|
|
# isFuzzed - whether this message should have its subcomponent
|
|
# flag isFuzzed set
|
|
def setMessageFrom(self, sourceType, message, isFuzzed):
|
|
if sourceType == self.Format.CommaSeparatedHex:
|
|
message = bytearray([x.decode("hex") for x in message.split(",")])
|
|
elif sourceType == self.Format.Ascii:
|
|
message = self.deserializeByteArray(message)
|
|
elif sourceType == self.Format.Raw:
|
|
message = message
|
|
else:
|
|
raise RuntimeError("Invalid sourceType")
|
|
|
|
self.subcomponents = [MessageSubComponent(message, isFuzzed)]
|
|
|
|
if isFuzzed:
|
|
self.isFuzzed = True
|
|
|
|
# Same arguments as above, but adds to .message as well as
|
|
# adding a new subcomponent
|
|
# createNewSubcomponent - If false, don't create another subcomponent,
|
|
# instead, append new message data to last subcomponent in message
|
|
def appendMessageFrom(self, sourceType, message, isFuzzed, createNewSubcomponent=True):
|
|
if sourceType == self.Format.CommaSeparatedHex:
|
|
newMessage = bytearray([x.decode("hex") for x in message.split(",")])
|
|
elif sourceType == self.Format.Ascii:
|
|
newMessage = self.deserializeByteArray(message)
|
|
elif sourceType == self.Format.Raw:
|
|
newMessage = message
|
|
else:
|
|
raise RuntimeError("Invalid sourceType")
|
|
|
|
if createNewSubcomponent:
|
|
self.subcomponents.append(MessageSubComponent(newMessage, isFuzzed))
|
|
else:
|
|
self.subcomponents[-1].message += newMessage
|
|
|
|
if isFuzzed:
|
|
# Make sure message is set to fuzz as well
|
|
self.isFuzzed = True
|
|
|
|
def isOutbound(self):
|
|
return self.direction == self.Direction.Outbound
|
|
|
|
def __eq__(self, other):
|
|
# bytearray (for message) implements __eq__()
|
|
return self.direction == other.direction and self.message == other.message
|
|
|
|
@classmethod
|
|
def serializeByteArray(cls, byteArray):
|
|
return repr(str(byteArray))
|
|
|
|
@classmethod
|
|
def deserializeByteArray(cls, string):
|
|
# This appears to properly reverse repr() without the risks of eval
|
|
return bytearray(string[1:-1].encode('utf8').decode('unicode-escape').encode('utf8'))
|
|
|
|
def getAlteredSerialized(self):
|
|
if len(self.subcomponents) < 1:
|
|
return "{0} {1}\n".format(self.direction, "ERROR: No data in message.")
|
|
else:
|
|
serializedMessage = "{0}{1} {2}\n".format("fuzz " if self.subcomponents[0].isFuzzed else "", self.direction, self.serializeByteArray(self.subcomponents[0].getAlteredByteArray()))
|
|
|
|
for subcomponent in self.subcomponents[1:]:
|
|
serializedMessage += "sub {0}{1}\n".format("fuzz " if subcomponent.isFuzzed else "", self.serializeByteArray(subcomponent.getAlteredByteArray()))
|
|
|
|
return serializedMessage
|
|
|
|
def getSerialized(self):
|
|
if len(self.subcomponents) < 1:
|
|
return "{0} {1}\n".format(self.direction, "ERROR: No data in message.")
|
|
else:
|
|
serializedMessage = "{0} {1}{2}\n".format(self.direction, "fuzz " if self.subcomponents[0].isFuzzed else "", self.serializeByteArray(self.subcomponents[0].message))
|
|
|
|
for subcomponent in self.subcomponents[1:]:
|
|
serializedMessage += "sub {0}{1}\n".format("fuzz " if subcomponent.isFuzzed else "", self.serializeByteArray(subcomponent.message))
|
|
|
|
return serializedMessage
|
|
|
|
# Utility function for setFromSerialized and appendFromSerialized below
|
|
def _extractMessageComponents(self, serializedData):
|
|
firstQuoteSingle = serializedData.find('\'')
|
|
lastQuoteSingle = serializedData.rfind('\'')
|
|
firstQuoteDouble = serializedData.find('"')
|
|
lastQuoteDouble = serializedData.rfind('"')
|
|
firstQuote = -1
|
|
lastQuote = -1
|
|
|
|
if firstQuoteSingle == -1 or firstQuoteSingle == lastQuoteSingle:
|
|
# If no valid single quotes, go double quote
|
|
firstQuote = firstQuoteDouble
|
|
lastQuote = lastQuoteDouble
|
|
elif firstQuoteDouble == -1 or firstQuoteDouble == lastQuoteDouble:
|
|
# If no valid double quotes, go single quote
|
|
firstQuote = firstQuoteSingle
|
|
lastQuote = lastQuoteSingle
|
|
elif firstQuoteSingle < firstQuoteDouble:
|
|
# If both are valid, go single if further out
|
|
firstQuote = firstQuoteSingle
|
|
lastQuote = lastQuoteSingle
|
|
else:
|
|
# Both are valid but double is further out
|
|
firstQuote = firstQuoteDouble
|
|
lastQuote = lastQuoteDouble
|
|
|
|
if firstQuote == -1 or lastQuote == -1 or firstQuote == lastQuote:
|
|
raise RuntimeError("Invalid message data, no message found")
|
|
|
|
# Pull out everything, quotes and all, and deserialize it
|
|
messageData = serializedData[firstQuote:lastQuote+1]
|
|
# Process the args
|
|
serializedData = serializedData[:firstQuote].split(" ")
|
|
|
|
return (serializedData, messageData)
|
|
|
|
# Handles _one line_ of data, either "inbound" or "outbound"
|
|
# Lines following this should be passed to appendFromSerialized() below
|
|
def setFromSerialized(self, serializedData):
|
|
serializedData = serializedData.replace("\n", "")
|
|
(serializedData, messageData) = self._extractMessageComponents(serializedData)
|
|
|
|
if len(messageData) == 0 or len(serializedData) < 1:
|
|
raise RuntimeError("Invalid message data")
|
|
|
|
direction = serializedData[0]
|
|
args = serializedData[1:-1]
|
|
|
|
if direction != "inbound" and direction != "outbound":
|
|
raise RuntimeError("Invalid message data, unknown direction {0}".format(direction))
|
|
|
|
isFuzzed = False
|
|
if "fuzz" in args:
|
|
isFuzzed = True
|
|
if len(serializedData) < 3:
|
|
raise RuntimeError("Invalid message data")
|
|
|
|
self.direction = direction
|
|
self.setMessageFrom(self.Format.Ascii, messageData, isFuzzed)
|
|
|
|
# Add another line, used for multiline messages
|
|
def appendFromSerialized(self, serializedData, createNewSubcomponent=True):
|
|
serializedData = serializedData.replace("\n", "")
|
|
(serializedData, messageData) = self._extractMessageComponents(serializedData)
|
|
|
|
if createNewSubcomponent:
|
|
if len(messageData) == 0 or len(serializedData) < 1 or serializedData[0] != "sub":
|
|
raise RuntimeError("Invalid message data")
|
|
else:
|
|
# If not creating a subcomponent, we won't have "sub", "fuzz", and the other fun stuff
|
|
if len(messageData) == 0:
|
|
raise RuntimeError("Invalid message data")
|
|
|
|
args = serializedData[1:-1]
|
|
# Put either "fuzz" or nothing before actual message
|
|
# Can tell the difference even with ascii because ascii messages have '' quotes
|
|
# IOW, even a message subcomponent 'fuzz' will have the 's around it, not be fuzz without quotes
|
|
isFuzzed = False
|
|
if "fuzz" in args:
|
|
isFuzzed = True
|
|
|
|
self.appendMessageFrom(self.Format.Ascii, messageData, isFuzzed, createNewSubcomponent=createNewSubcomponent)
|
|
|
|
class MessageCollection(object):
|
|
def __init__(self):
|
|
self.messages = []
|
|
|
|
def addMessage(self, message):
|
|
self.messages.append(message)
|
|
|
|
def doClientMessagesMatch(self, otherMessageCollection):
|
|
for i in range(0, len(self.messages)):
|
|
# Skip server messages
|
|
if not self.messages[i].isOutbound():
|
|
continue
|
|
try:
|
|
# Message implements __eq__()
|
|
if self.messages[i] != otherMessageCollection.messages[i]:
|
|
return False
|
|
except IndexError:
|
|
return False
|
|
|
|
# All messages passed
|
|
return True
|
|
|
|
import os
|
|
import os.path
|
|
from copy import deepcopy
|
|
|
|
# Handles all the logging of the fuzzing session
|
|
# Log messages can be found at sample_apps/<app>/<app>_logs/<date>/
|
|
class Logger(object):
|
|
def __init__(self, folderPath):
|
|
self._folderPath = folderPath
|
|
if os.path.exists(folderPath):
|
|
print("Data output directory already exists: %s" % (folderPath))
|
|
exit()
|
|
else:
|
|
try:
|
|
os.makedirs(folderPath)
|
|
except:
|
|
print("Unable to create logging directory: %s" % (folderPath))
|
|
exit()
|
|
|
|
self.resetForNewRun()
|
|
|
|
# Store just the data, forget trying to make a Message object
|
|
# With the subcomponents and everything, it just gets weird,
|
|
# and we don't need it
|
|
def setReceivedMessageData(self, messageNumber, data):
|
|
self.receivedMessageData[messageNumber] = data
|
|
|
|
def setHighestMessageNumber(self, messageNumber):
|
|
# The highest message # this fuzz session made it to
|
|
self._highestMessageNumber = messageNumber
|
|
|
|
def outputLastLog(self, runNumber, messageCollection, errorMessage):
|
|
return self._outputLog(runNumber, messageCollection, errorMessage, self._lastReceivedMessageData, self._lastHighestMessageNumber)
|
|
|
|
def outputLog(self, runNumber, messageCollection, errorMessage):
|
|
return self._outputLog(runNumber, messageCollection, errorMessage, self.receivedMessageData, self._highestMessageNumber)
|
|
|
|
def _outputLog(self, runNumber, messageCollection, errorMessage, receivedMessageData, highestMessageNumber):
|
|
with open(os.path.join(self._folderPath, str(runNumber)), "w") as outputFile:
|
|
print("Logging run number %d" % (runNumber))
|
|
outputFile.write("Log from run with seed %d\n" % (runNumber))
|
|
outputFile.write("Error message: %s\n" % (errorMessage))
|
|
|
|
if highestMessageNumber == -1 or runNumber == 0:
|
|
outputFile.write("Failed to connect on this run.\n")
|
|
|
|
outputFile.write("\n")
|
|
|
|
i = 0
|
|
for message in messageCollection.messages:
|
|
outputFile.write("Packet %d: %s" % (i, message.getSerialized()))
|
|
|
|
if message.isFuzzed:
|
|
outputFile.write("Fuzzed Packet %d: %s\n" % (i, message.getAlteredSerialized()))
|
|
|
|
if i in receivedMessageData:
|
|
# Compare what was actually sent to what we expected, log if they differ
|
|
if receivedMessageData[i] != message.getOriginalMessage():
|
|
outputFile.write("Actual data received for packet %d: %s" % (i, Message.serializeByteArray(receivedMessageData[i])))
|
|
else:
|
|
outputFile.write("Received expected data\n")
|
|
|
|
if highestMessageNumber == i:
|
|
if message.isOutbound():
|
|
outputFile.write("This is the last message sent\n")
|
|
else:
|
|
outputFile.write("This is the last message received\n")
|
|
|
|
outputFile.write("\n")
|
|
i += 1
|
|
|
|
def resetForNewRun(self):
|
|
try:
|
|
self._lastReceivedMessageData = deepcopy(self.receivedMessageData)
|
|
self._lastHighestMessageNumber = self._highestMessageNumber
|
|
except AttributeError:
|
|
self._lastReceivedMessageData = {}
|
|
self._lastHighestMessageNumber = -1
|
|
|
|
self.receivedMessageData = {}
|
|
self.setHighestMessageNumber(-1)
|