Script smb-psexec

Script types: hostrule
Categories: intrusive

Script Summary

Implements remote process execution similar to the Sysinternals' psexec tool, allowing a user to run a series of programs on a remote machine and read the output. This is great for gathering information about servers, running the same tool on a range of system, or even installing a backdoor on a collection of computers.

This script can run commands present on the remote machine, such as ping or tracert, or it can upload a program and run it, such as pwdump6 or a backdoor. Additionally, it can read the program's stdout/stderr and return it to the user (works well with ping, pwdump6, etc), or it can read a file that the process generated (fgdump, for example, generates a file), or it can just start the process and let it run headless (a backdoor might run like this).

To use this, a configuration file should be created and edited. Several configuration files are included that you can customize, or you can write your own. This config file is placed in nselib/data/psexec (if you aren't sure where that is, search your system for default.lua), then is passed to Nmap as a script argument (for example, myconfig.lua would be passed as --script-args=config=myconfig.

The configuration file consists mainly of a module list. Each module is defined by a lua table, and contains fields for the name of the program, the executable and arguments for the program, and a score of other options. Modules also have an 'upload' field, which determines whether or not the module is to be uploaded. Here is a simple example of how to run

localgroup administrators
, which returns a list of users in the "administrators" group (take a look at the examples.lua configuration file for these examples):

  mod = {}
  mod.upload           = false             = "Example 1: Membership of 'administrators'"
  mod.program          = "net.exe"
  mod.args             = "localgroup administrators"
  table.insert(modules, mod)

mod.upload is false, meaning the program should already be present on the remote system (since 'net.exe' is on every version of Windows, this should be the case). defines the name that the program will have in the output. mod.program and mod.args obviously define which program is going to be run. The output for this script is this:

  |  Example 1: Membership of 'administrators'
  |  | Alias name     administrators
  |  | Comment        Administrators have complete and unrestricted access to the computer/domain
  |  |
  |  | Members
  |  |
  |  | -------------------------------------------------------------------------------
  |  | Administrator
  |  | ron
  |  | test
  |  | The command completed successfully.
  |  |
  |  |_

That works, but it's really ugly. In general, we can use mod.find, mod.replace, mod.remove, and mod.noblank to clean up the output. For this example, we're going to use mod.remove to remove a lot of the useless lines, and mod.noblank to get rid of the blank lines that we don't want:

  mod = {}
  mod.upload           = false             = "Example 2: Membership of 'administrators', cleaned"
  mod.program          = "net.exe"
  mod.args             = "localgroup administrators"
  mod.remove           = {"The command completed", "%-%-%-%-%-%-%-%-%-%-%-", "Members", "Alias name", "Comment"}
  mod.noblank          = true
  table.insert(modules, mod)

We can see that the output is now much cleaner:

|  Example 2: Membership of 'administrators', cleaned
|  | Administrator
|  | ron
|  |_test

For our next command, we're going to run Windows' ipconfig.exe, which outputs a significant amount of unnecessary information, and what we do want isn't formatted very nicely. All we want is the IP address and MAC address, and we get it using mod.find and mod.replace:

  mod = {}
  mod.upload           = false             = "Example 3: IP Address and MAC Address"
  mod.program          = "ipconfig.exe"
  mod.args             = "/all"
  mod.maxtime          = 1
  mod.find             = {"IP Address", "Physical Address", "Ethernet adapter"}
  mod.replace          = {{"%. ", ""}, {"-", ":"}, {"Physical Address", "MAC Address"}}
  table.insert(modules, mod)

This module searches for lines that contain "IP Address", "Physical Address", or "Ethernet adapter". In these lines, a ". " is replaced with nothing, a "-" is replaced with a colon, and the term "Physical Address" is replaced with "MAC Address" (arguably unnecessary). Run ipconfig /all yourself to see what we start with, but here's the final output:

|  Example 3: IP Address and MAC Address
|  | Ethernet adapter Local Area Connection:
|  |    MAC Address: 00:0C:29:12:E6:DB
|  |_   IP Address:|  Example 3: IP Address and MAC Address

Another interesting part of this script is that variables can be used in any script fields. There are two types of variables: built-in and user-supplied. Built-in variables can be anything found in the config table, most of which are listed below. The more interesting ones are:

  • $lhost: The address of the scanner
  • $rhost: The address being scanned
  • $path: The path where the scripts are uploaded
  • $share: The share where the script was uploaded

User-supplied arguments are given on the commandline, and can be controlled by mod.req_args in the configuration file. Arguments are given by the user in --script-args; for example, to set $host to '', the user would pass in --script-args=host= To ensure the user passes in the host variable, mod.req_args would be set to {'host'}.

Here is a module that pings the local ip address:

  mod = {}
  mod.upload           = false             = "Example 4: Can the host ping our address?"
  mod.program          = "ping.exe"
  mod.args             = "$lhost"
  mod.remove           = {"statistics", "Packet", "Approximate", "Minimum"}
  mod.noblank          = true
  mod.env              = "SystemRoot=c:\\WINDOWS"
  table.insert(modules, mod)

And the output:

|  Example 4: Can the host ping our address?
|  | Pinging with 32 bytes of data:
|  | Reply from bytes=32 time<1ms TTL=64
|  | Reply from bytes=32 time<1ms TTL=64
|  | Reply from bytes=32 time<1ms TTL=64
|  |_Reply from bytes=32 time<1ms TTL=64

And this module pings an arbitrary address that the user is expected to give:

  mod = {}
  mod.upload           = false             = "Example 5: Can the host ping $host?"
  mod.program          = "ping.exe"
  mod.args             = "$host"
  mod.remove           = {"statistics", "Packet", "Approximate", "Minimum"}
  mod.noblank          = true
  mod.env              = "SystemRoot=c:\\WINDOWS"
  mod.req_args         = {'host'}
  table.insert(modules, mod)

And the output (note that we had to up the timeout so this would complete; we'll talk about override values later):

$ ./nmap -n -d -p445 --script=smb-psexec --script-args=smbuser=test,smbpass=test,config=examples,host=
|  Example 5: Can the host ping
|  | Pinging with 32 bytes of data:
|  | Request timed out.
|  | Request timed out.
|  | Request timed out.
|  |_Request timed out.

For the final example, we'll use the upload command to upload fgdump.exe, run it, download its output file, and clean up its logfile. You'll have to put fgdump.exe in the same folder as the script for this to work:

  mod = {}
  mod.upload           = true             = "Example 6: FgDump"
  mod.program          = "fgdump.exe"
  mod.args             = "-c -l fgdump.log"
  mod.url              = ""
  mod.tempfiles        = {"fgdump.log"}
  mod.outfile          = ""
  table.insert(modules, mod)

The -l argument for fgdump supplies the name of the logfile. That file is listed in the mod.tempfiles field. What, exactly, does mod.tempfiles do? It simply gives the service a list of files to delete while cleaning up. The cleanup process will be discussed later.

mod.url is displayed to the user if mod.program isn't found in nselib/data/psexec/. And finally, mod.outfile is the file that is downloaded from the system. This is required because fgdump writes to an output file instead of to stdout (pwdump6, for example, doesn't require mod.outfile.

Now that we've seen a few possible combinations of fields, I present a complete list of all fields available and what each of them do. Many of them will be familiar, but there are a few that aren't discussed in the examples:

  • upload (boolean) true if it's a local file to upload, false if it's already on the host machine. If upload is true, program has to be in nselib/data/psexec.
  • name (string) The name to display above the output. If this isn't given, program .. args are used.
  • program (string) If upload is false, the name (fully qualified or relative) of the program on the remote system; if upload is true, the name of the local file that will be uploaded (stored in nselib/data/psexec).
  • args (string) Arguments to pass to the process.
  • env (string) Environmental variables to pass to the process, as name=value pairs, delimited, per Microsoft's spec, by NULL characters (string.char(0)).
  • maxtime (integer) The approximate amount of time to wait for this process to complete. The total timeout for the script before it gives up waiting for a response is the total of all maxtime fields.
  • extrafiles (string[]) Extra file(s) to upload before running the program. These will not be renamed (because, presumably, if they are then the program won't be able to find them), but they will be marked as hidden/system/etc. This may cause a race condition if multiple people are doing this at once, but there isn't much we can do. The files are also deleted afterwards as tempfiles would be. The files have to be in the same directory as programs (nselib/data/psexec), but the program doesn't necessarily need to be an uploaded one.
  • tempfiles (string[]) A list of temporary files that the process is known to create (if the process does create files, using this field is recommended because it helps avoid making a mess on the remote system).
  • find (string[]) Only display lines that contain the given string(s) (for example, if you're searching for a line that contains "IP Address", set this to
    . This allows Lua-style patterns, see: (don't forget to escape special characters with a %). Note that this is client-side only; the full output is still returned, the rest is removed while displaying. The line of output only needs to match one of the strings given here.
  • remove (string[]) Opposite of find; this removes lines containing the given string(s) instead of displaying them. Like find, this is client-side only and uses Lua-style patterns. If remove and find are in conflict, then remove takes priority.
  • noblank (boolean) Setting this to true removes all blank lines from the output.
  • replace (table) A table of values to replace in the strings returned. Like find and replace, this is client-side only and uses Lua-style patterns.
  • headless (boolean) If headless is set to true, the program doesn't return any output; rather, it runs detached from the service so that, when the service ends, the program keeps going. This can be useful for, say, a monitoring program. Or a backdoor, if that's what you're into (a Metasploit payload should work nicely). Not compatible with: find, remove, noblank, replace, maxtime, outfile.
  • enabled (boolean) Set to false, and optionally set disabled_message, if you don't want a module to run. Alternatively, you can comment out the process.
  • disabled_message (string) Displayed if the module is disabled.
  • url (string) A module where the user can download the uploadable file. Displayed if the uploadable file is missing.
  • outfile (string) If set, the specified file will be returned instead of stdout.
  • req_args (string[]) An array of arguments that the user must set in --script-args.

Any field in the configuration file can contain variables, as discussed. Here are some of the available built-in variables:

  • $lhost: local IP address as a string.
  • $lport: local port (meaningless; it'll change by the time the module is uploaded since multiple connections are made).
  • $rhost: remote IP address as a string.
  • $rport: remote port.
  • $lmac: local MAC address as a string in the xx:xx:xx:xx:xx:xx format (note: requires root).
  • $path: the path where the file will be uploaded to.
  • $service_name: the name of the service that will be running this program
  • $service_file: the name of the executable file for the service
  • $temp_output_file: The (ciphered) file where the programs' output will be written before being renamed to $output_file
  • $output_file: The final name of the (ciphered) output file. When this file appears, the script downloads it and stops the service
  • $timeout: The total amount of time the script is going to run before it gives up and stops the process
  • $share: The share that everything was uploaded to
  • (script args): Any value passed as a script argument will be replaced (for example, if Nmap is run with --script-args=var3=10, then $var3 in any field will be replaced with 10. See the req_args field above. Script argument values take priority over config values.

In addition to modules, the configuration file can also contain overrides. Most of these aren't useful, so I'm not going to go into great detail. Search smb-psexec.nse for any reference to the config table; any value in the config table can be overridden with the overrides table in the module. The most useful value to override is probably timeout.

Before and after scripts are run, and when there's an error, a cleanup is performed. in the cleanup, we attempt to stop the remote processes, delete all programs, output files, temporary files, extra files, etc. A lot of effort was put into proper cleanup, since making a mess on remote systems is a bad idea.

Now that I've talked at length about how to use this script, I'd like to spend some time talking about how it works.

Running a script happens in several stages:

  1. An open fileshare is found that we can write to. Finding an open fileshare basically consists of enumerating all shares and seeing which one(s) we have access to.
  2. A "service wrapper", and all of the uploadable/extra files, are uploaded. Before they're uploaded, the name of each file is obfuscated. The obfuscation completely renames the file, is unique for each source system, and doesn't change between multiple runs. This obfuscation has the benefit of preventing filenames from overlapping if multiple people are running this against the same computer, and also makes it more difficult to determine their purposes. The reason for keeping them consistent for every run is to make cleanup possible: a random filename, if the script somehow fails, will be left on the system.
  3. A new service is created and started. The new service has a random name for the same reason the files do, and points at the 'service wrapper' program that was uploaded.
  4. The service runs the processes. One by one, the processes are run and their output is captured. The output is obfuscated using a simple (and highly insecure) xor algorithm, which is designed to prevent casual sniffing (but won't deter intelligent attackers). This data is put into a temporary output file. When all the programs have finished, the file is renamed to the final output file
  5. The output file is downloaded, and the cleanup is performced. The file being renamed triggers the final stage of the program, where the data is downloaded and all relevant files are deleted.
  6. Output file, now decrypted, is formatted and displayed to the user.

And that's how it works!

Please post any questions, or suggestions for better modules, to

And, as usual, since this tool can be dangerous and can easily be viewed as a malicious tool -- use this responsibly, and don't break any laws with it.

Some ideas for later versions (TODO):

  • Set up a better environment for scripts (PATH, SystemRoot, etc). Without this, a lot of programs (especially ones that deal with network traffic) behave oddly.
  • Abstract the code required to run remote processes so other scripts can use it more easily (difficult, but will ultimately be well worth it later). (May actually not be possible. There is a lot of overhead and specialized code in this module. We'll see, though.)
  • Let user specify an output file (per-script) so they can, for example, download binary files (don't think it's worthwhile).
  • Consider running the external programs in parallel (not sure if the benefits outweigh the drawbacks).
  • Let the config request the return code from the process instead of the output (not sure if doing this would be worth the effort).
  • Check multiple shares in a single session to save packets (and see where else we can tighten up the amount of traffic).

Script Arguments


Don't set the uploaded files to hidden/system/etc.


Set to only clean up any mess we made (leftover files, processes, etc. on the host OS) on a previous run of the script. This will attempt to delete the files from every share, not just the first one. This is done to prevent leftover files if the OS changes the ordering of the shares (there's no guarantee of shares coming back in any particular order) Note that cleaning up is still fairly invasive, since it has to re-discover the proper share, connect to it, delete files, open the services manager, etc.


Set to disable the ciphering of the returned text (useful for debugging).


The full path to the share (eg, "c:\windows"). This is required when creating a service.


The config file to use (eg, default). Config files require a .lua extension, and are located in nselib/data/psexec.


The minimum amount of time, in seconds, to wait for the external module to finish (default: 15)


Set to not clean up at all; this leaves the files on the remote system and the wrapper service installed. This is bad in practice, but significantly reduces the network traffic and makes analysis easier.


Script uses this value instead of a random encryption key (useful for debugging the crypto).


Set to override the share used for uploading. This also stops shares from being enumerated, and all other shares will be ignored. No checks are done to determine whether or not this is a valid share before using it. Reqires sharepath to be set.

randomseed, smbbasic, smbport, smbsign

See the documentation for the smb library.

smbdomain, smbhash, smbnoguest, smbpassword, smbtype, smbusername

See the documentation for the smbauth library.

Example Usage

nmap --script smb-psexec.nse --script-args=smbuser=<username>,smbpass=<password>[,config=<config>] -p445 <host>
sudo nmap -sU -sS --script smb-psexec.nse --script-args=smbuser=<username>,smbpass=<password>[,config=<config>] -p U:137,T:139 <host>

Script Output

Host script results:
|  smb-psexec:
|  |  Windows version
|  |  |_ Microsoft Windows 2000 [Version 5.00.2195]
|  |  IP Address and MAC Address from 'ipconfig.exe'
|  |  |  Ethernet adapter Local Area Connection 2:
|  |  |         MAC Address: 00:50:56:A1:24:C2
|  |  |         IP Address:
|  |  |  Ethernet adapter Local Area Connection:
|  |  |_        MAC Address: 00:50:56:A1:00:65
|  |  User list from 'net user'
|  |  |  Administrator            TestUser3                Guest
|  |  |  IUSR_RON-WIN2K-TEST      IWAM_RON-WIN2K-TEST      nmap
|  |  |  rontest123               sshd                     SvcCOPSSH
|  |  |_ test1234                 Testing                  TsInternetUser
|  |  Membership of 'administrators' from 'net localgroup administrators'
|  |  |  Administrator
|  |  |  SvcCOPSSH
|  |  |  test1234
|  |  |_ Testing
|  |  Can the host ping our address?
|  |  |  Pinging with 32 bytes of data:
|  |  |_ Reply from bytes=32 time<10ms TTL=64
|  |  Traceroute back to the scanner
|  |  |_   1   <10 ms   <10 ms   <10 ms
|  |  ARP Cache from arp.exe
|  |  |    Internet Address      Physical Address      Type
|  |  |_            00-50-56-a1-27-4b     dynamic
|  |  List of listening and established connections (netstat -an)
|  |  |    Proto  Local Address          Foreign Address        State
|  |  |    TCP                 LISTENING
|  |  |    TCP                 LISTENING
|  |  |    TCP                 LISTENING
|  |  |    TCP                LISTENING
|  |  |    TCP                LISTENING
|  |  |    TCP                LISTENING
|  |  |    TCP               LISTENING
|  |  |    TCP               LISTENING
|  |  |    TCP               LISTENING
|  |  |    TCP               LISTENING
|  |  |    TCP               LISTENING
|  |  |    TCP              LISTENING
|  |  |    TCP         ESTABLISHED
|  |  |    TCP         ESTABLISHED
|  |  |    TCP         ESTABLISHED
|  |  |    TCP         ESTABLISHED
|  |  |    TCP              LISTENING
|  |  |    TCP         CLOSE_WAIT
|  |  |    UDP            *:*
|  |  |    UDP            *:*
|  |  |    UDP           *:*
|  |  |    UDP           *:*
|  |  |    UDP          *:*
|  |  |    UDP          *:*
|  |  |    UDP          *:*
|  |  |    UDP         *:*
|  |  |_   UDP         *:*
|  |  Full routing table from 'netstat -nr'
|  |  |  ===========================================================================
|  |  |  Interface List
|  |  |  0x1 ........................... MS TCP Loopback interface
|  |  |  0x2 ...00 50 56 a1 00 65 ...... VMware Accelerated AMD PCNet Adapter
|  |  |  0x1000004 ...00 50 56 a1 24 c2 ...... VMware Accelerated AMD PCNet Adapter
|  |  |  ===========================================================================
|  |  |  ===========================================================================
|  |  |  Active Routes:
|  |  |  Network Destination        Netmask          Gateway       Interface  Metric
|  |  |       1
|  |  |      1
|  |  |      1
|  |  |      1
|  |  |      1
|  |  |               2      1
|  |  |  ===========================================================================
|  |  |  Persistent Routes:
|  |  |    None
|_ |_ |_ Route Table



  • Ron Bowes

License: Same as Nmap--See