- classes.dex gets mangled on windows; use binary mode when reading it
- UnknownHostExceptions on API Level 3 emulator because of trailing
whitespace after the hostname/IP
- Work around integer overflow at year 2038 when signing the payload
Convert the dx calls from build.sh to equivalent exec calls in Maven
deploy profile.
While this commit takes into account differences between Windows and *nix,
it was only tested on Windows, and the resulting binaries have not been
tested at all!
In addition, I was not able to pass individual .class file names to dx
without getting a "class name does not match path" error, so I changed it
to copy all required classes into a temp directory and call dx from there.
I also changed the cross-project paths to refer to the respective Maven
classpath, so in case you do an individual project build, the library
dependencies are taken from the Maven repository instead of taking them
from the target/ directory of the projects directly.
Include the animal-scents for Android API in this commit, so that users
who do not have Android SDK can still check meterpreter API compatibility
with Android API. Some classes, like screenshot have been excluded since
they need AWT (but they are excluded in Android Meterpreter anyway).
To regenerate the scents file, run
mvn -Dandroid.sdk.path=... -P regenerate package
Apparently Android API 3 does not know the getMTU() function, which was
added in Java 1.6, and in Android API Level 9 (Gingerbread). Therefore,
fall back to the 1.4 version that does not need this API.
Apparently, getNetworkPrefixLength can return -1, which confuses the Ruby
side. Therefore fall back to guessing the prefix in this case, as we do it
for Java <= 1.6.
Apparently, my last update came from the future. This modification
to that future update fixes an oversight preventing Armitage from
connecting to its collaboration server because it would report the
wrong application.
This update to Armitage improves its responsiveness when connected
to a team server over a high latency network. This update also adds
a publish/query/subscribe API to Cortana.
The BindTCP test contained a race condition: if the bind payload took
longer to load than the handler, it could result in a
ConnectException: Connection refused: connect
Work around this by retrying the connection up to 10 times, with 500ms
delay in between.
Due to the modular structure of payload stages its pretty trivial
to add DNS resolution instead of hard-coded IP address in stage0.
The only real complication here is that ReverseConnectRetries ends
up being one byte further down than in the original shellcode. It
appears that the original rev_tcp_dns payload suffers from the same
issue.
Hostname substitution is handled in the same method as the RC4 and
XOR keys, with an offset provided and replace_vars ignoring the
hostname.
Tested in x86 native and WOW64 on XP and 2k8r2 respectively.
This is a good option for those of us needing to leave persistent
binaries/payloads on hosts for long periods. Even if the hostname
resolves to a malicious party attempting to steal our hard earned
session, they'd be hard pressed to crypt the payload with the
appropriate RC4 pass. So long as we control the NS and records, the
hardenned shellcode should provide a better night's sleep if running
shells over the WAN. Changing the RC4 password string in the
shellcode and build.py should reduce the chances of recovery by RE.
Next step will likely be to start generating elipses for ECDH SSL
in meterpreter sessions and passing them with stage2 through the
RC4 socket. If P is 768-1024 the process is relatively quick, but
we may want to precompute a few defaults as well to have 2048+.
The HttpOpenRequest function from WinINet requires the
INTERNET_FLAG_KEEP_CONNECTION flag to communicate through an
authenticated proxy.
From MSDN ( http://tinyurl.com/chwt86j ):
"Uses keep-alive semantics, if available, for the connection. This
flag is required for Microsoft Network (MSN), NT LAN Manager (NTLM),
and other types of authentication."
Without this flag, the HTTP stager will fail when faced with a proxy
that requires authentication. The Windows HTTPS stager does not have
this problem.
For HTTP Meterpreter to communicate through an authenticated proxy a
separate patch will need to be made to the Meterpreter source code.
This is at line 1125 of source/common/core.c in the Meterpreter source
code.
My motivation for this request is for windows/dllinject/reverse_http
to download a DLL even when faced with an authenticated proxy. These
changes accomplish this.
Test environment:
I staged a SmoothWall device with the Advanced Proxy Web Add-on. I
enabled Integrated Windows Authentication with a W2K3 DC. I verified
the HTTP stager authenticated to and communicated through the proxy
by watching the proxy access.log
This update adds the ability to manage multiple team server instances
through one Armitage client. This update also adds nickname completion
to the event log. Several bug fixes are included too.
This update to Armitage adds the ability to assign labels to hosts
and create dynamic workspaces based on these labs. This update also
adds helpers to configure USERNAME/PASSWORD options and EXE::Custom
and EXE::Template. Several bugs were fixed as well.
Armitage on Windows requires the user to specify their MSF
install folder. This tweak checks for an MSF 4.5 environment
and updates the specified folder to make everything work.
Like magic.
This is Armitage release 01.04.13. This update fixes several bugs
and improves the user experience launching *_login modules from
Armitage. This update adds a Windows 8 icon and includes a fix to
better work with the Metasploit 1.45 installer's environment.
ESI is not clobbered; no need to clear EDX as only DL is filled before and
it is overwritten before use.
Shellcodes in ruby modules not regenerated, but I guess you want to
regenerate them again anyway :-)
Those stagers will encrypt the initial stage with a 128-bit RC4 key and
the stage length with a XOR key. Both keys are embedded in the stager.
This should provide good evasion capabilities in addition to some
protection against MITM reversing (if the stager is sent a different
route, like in an executable on an USB key).
Note that, from a cryptanalyst's standpoint, it is a bad idea to reuse the
same stager (or stagers with the same RC4 and XOR keys) more than once
since an identical key will result in an identical keystream and make
correlation attacks easy. But I doubt that matters in practice.
Also note that since communication after the initial statging is not
encrypted, these stagers should be used in combination with additional
encryption support in the payloads (like Meterpreter).
Provided as a block to be included into stagers and/or decoder stubs.
Also included is a test shellcode that can be used for verifying that the
algorithm is compatible to Ruby's OpenSSL RC4 algorithm.
Downgrad JUnit version since JUnit 4 can only work with -target 1.5 or
higher class files.
Covered are Shell and Meterpreter stage, StreamForwarder, MemoryBuffer,
AESEncryption and Payload (Bind, Reverse, Spawn, AESPassword).
Check JavaPayload and Java Meterpreter against version incompatibilities
for Java 1.2, 1.3, 1.4, 1.5, and 1.6.
Note that webcam_audio_record is currently excluded from the checks, as it
uses Sun proprietary API for building the WAV file and is therefore
failing the build (and will most likely crash Meterpreter if run on a JVM
of version 1.4 or later that is not based on the Sun/Oracle JVM).
Possible workarounds (apart from either removing the module or changing it
to produce empty files when WAV creation is not supported) include
implementing the WAV file writer ourselves or providing raw PCM files
instead.
- Adds a call to mprotect(2) to the reverse and bind stagers
- Adds accurate source for some other linux shellcode, including some
comments to make it more maintainable
- Adds tools/module_payload.rb for listing all payloads for each exploit
in a greppable format. Makes it easy to find out if a payload change
causes a payload to no longer be compatible with a given exploit.
- Missing from this commit is source for reverse_ipv6_tcp
Now you can "make single_bind_tcp_shell", or the like, and build one
payload instead of the kludgy embedded shell script that always builds
all of them.
Need to do the same with BSD.
Instead of invoking the Watson crashamajigger when the process
associated with a connection has a long executable name, truncate to the
length available in the buffer.
[See #609]
Fixes client.sys.process.execute for posix, which previously (since
2010!) would always return nil, or a single byte. This makes sense
considering the value of bytesRead would always be either 0 or 1 because
it was being assigned the result of the comparison instead of the return
value of read().
[Fixes#681]
Fixes some TypeError exceptions when attempting most operations on
spawned processes, e.g.:
p = client.sys.process.execute("/bin/sh", nil, "Channelized"=>true)
p.close
# raises TypeError: can't convert nil into Integer
[FIXRM #7005]
Squashed commit of the following:
commit 57795ff9c33a53167fca85845b96b82b5c92315f
Author: James Lee <egypt@metasploit.com>
Date: Wed Aug 1 14:13:20 2012 -0600
Add recompiled sniffer bin for linux
commit 0e11fdb06fcb9771a11eb631e6f10ec7a2d315f3
Author: m m <gaspmat@gmail.com>
Date: Thu Jul 12 15:08:10 2012 +0200
really limit packet count and data in linux sniffer
[Closes#605]
Ryan Wincey
Geyslan G. Bem
Meatballs
jvazquez-r7
Alexandre Maloteaux
corelanc0d3r
sinn3r
Matthias Kaiser
William Vu
James Lee
Tod Beardsley
agix
timwr
Michael Schierl
scriptjunkie
sinn3r
Raphael Mudge
RageLtMan
booboule
HD Moore
Tod Beardsley
m m
m m
LittleLightLittleFire
Stephen Fewer