ssh-audit/ssh-audit.1

.TH SSH-AUDIT 1 "April 18, 2024"
.SH NAME
\fBssh-audit\fP \- SSH server & client configuration auditor
.SH SYNOPSIS
.B ssh-audit
.RI [ options ] " <target_host>"
.SH DESCRIPTION
.PP
\fBssh-audit\fP analyzes the configuration of SSH servers & clients, then warns the user of weak, obsolete, and/or untested cryptographic primitives.  It is very useful for hardening SSH tunnels, which by default tend to be optimized for compatibility, not security.
.PP
See <https://www.ssh\-audit.com/> for official hardening guides for common platforms.

.SH OPTIONS
.TP
.B -h, \-\-help
.br
Print short summary of options.

.TP
.B -1, \-\-ssh1
.br
Only perform an audit using SSH protocol version 1.

.TP
.B -2, \-\-ssh2
.br
Only perform an audit using SSH protocol version 2.

.TP
.B -4, \-\-ipv4
.br
Prioritize the usage of IPv4.

.TP
.B -6, \-\-ipv6
.br
Prioritize the usage of IPv6.

.TP
.B -b, \-\-batch
.br
Enables grepable output.

.TP
.B -c, \-\-client\-audit
.br
Starts a server on port 2222 to audit client software configuration.  Use -p/--port=<port> to change port and -t/--timeout=<secs> to change listen timeout.

.TP
.B     \-\-conn\-rate\-test=N[:max_rate]
.br
Performs a connection rate test (useful for collecting metrics related to susceptibility of the DHEat vulnerability [CVE-2002-20001]).  A successful connection is counted when the server returns a valid SSH banner.  Testing is conducted with N concurrent sockets with an optional maximum rate of connections per second.

.TP
.B -d, \-\-debug
.br
Enable debug output.

.TP
.B     \-\-dheat=N[:kex[:e_len]]
.br
Run the DHEat DoS attack (CVE-2002-20001) against the target server (which will consume all available CPU resources).  The number of concurrent sockets, N, needed to achieve this effect will be highly dependent on the CPU resources available on the target, as well as the latency between the source and target machines.  The key exchange is automatically chosen based on which would cause maximum effect, unless explicitly chosen in the second field.  Lastly, an (experimental) option allows the length in bytes of the fake e value sent to the server to be specified in the third field.  Normally, the length of e is roughly the length of the modulus of the Diffie-Hellman exchange (hence, an 8192-bit / 1024-byte value of e is sent in each connection when targeting the diffie-hellman-group18-sha512 algorithm).  Instead, it was observed that many SSH implementations accept small values, such as 4 bytes; this results in a much more network-efficient attack.

.TP
.B -g, \-\-gex-test=<x[,y,...] | min1:pref1:max1[,min2:pref2:max2,...] | x-y[:step]>
.br
Runs a Diffie-Hellman Group Exchange modulus size test against a server.

Diffie-Hellman requires the client and server to agree on a generator value and a modulus value.  In the "Group Exchange" implementation of Diffie-Hellman, the client specifies the size of the modulus in bits by providing the server with minimum, preferred and maximum values. The server then finds a group that best matches the client's request, returning the corresponding generator and modulus.  For a full explanation of this process see RFC 4419 and its successors.

This test acts as a client by providing an SSH server with the size of a modulus and then obtains the size of the modulus returned by the server.

Three types of syntax are supported:

  1. <x[,y,...]>

     A comma delimited list of modulus sizes.
     A test is performed against each value in the list where it acts as the minimum, preferred and maximum modulus size.

  2. <min:pref:max[,min:pref:max,...]>

     A set of three colon delimited values denoting minimum, preferred and maximum modulus size.
     A test is performed against each set.
     Multiple sets can specified as a comma separated list.

  3. <x-y[:step]>

     A range of modulus sizes with an optional step value. Step defaults to 1 if omitted.
     If the left value is greater than the right value, then the sequence operates from right to left.
     A test is performed against each value in the range where it acts as the minimum, preferred and maximum modulus size.

Duplicates are excluded from the return value.

.TP
.B -j, \-\-json
.br
Output results in JSON format.  Specify twice (-jj) to enable indent printing (useful for debugging).

.TP
.B -l, \-\-level=<info|warn|fail>
.br
Specify the minimum output level.  Default is info.

.TP
.B -L, \-\-list-policies
.br
List all official, built-in policies for common systems.  Their full names can then be passed to -P/--policy.  Add \-v to \-L to view policy change logs.

.TP
.B \-\-lookup=<alg1,alg2,...>
.br
Look up the security information of an algorithm(s) in the internal database.  Does not connect to a server.

.TP
.B -m, \-\-manual
.br
Print the man page (Docker, PyPI, Snap, and Windows builds only).

.TP
.B -M, \-\-make-policy=<custom_policy.txt>
.br
Creates a policy based on the target server.  Useful when other servers should be compared to the target server's custom configuration (i.e.: a cluster environment).  Note that the resulting policy can be edited manually.

.TP
.B -n, \-\-no-colors
.br
Disable color output.  Automatically set when the NO_COLOR environment variable is set.

.TP
.B -p, \-\-port=<port>
.br
The TCP port to connect to when auditing a server, or the port to listen on when auditing a client.

.TP
.B -P, \-\-policy=<"built-in policy name" | path/to/custom_policy.txt>
.br
Runs a policy audit against a target using the specified policy (see \fBPOLICY AUDIT\fP section for detailed description of this mode of operation).  Combine with -c/--client-audit to audit a client configuration instead of a server.  Use -L/--list-policies to list all official, built-in policies for common systems.

.TP
.B     \-\-skip\-rate\-test
.br
Skips the connection rate test during standard audits.  By default, a few dozen TCP connections are created with the target host to see if connection throttling is implemented (this can safely infer whether the target is vulnerable to the DHEat attack; see CVE-2002-20001).

.TP
.B -t, \-\-timeout=<secs>
.br
The timeout, in seconds, for creating connections and reading data from the socket.  Default is 5.

.TP
.B -T, \-\-targets=<hosts.txt>
.br
A file containing a list of target hosts.  Each line must have one host, in the format of HOST[:PORT].  Use --threads to control concurrent scans.

.TP
.B     \-\-threads=<threads>
.br
The number of threads to use when scanning multiple targets (with -T/--targets).  Default is 32.

.TP
.B -v, \-\-verbose
.br
Enable verbose output.


.SH STANDARD AUDIT
.PP
By default, \fBssh-audit\fP performs a standard audit.  That is, it enumerates all host key types, key exchanges, ciphers, MACs, and other information, then color-codes them in output to the user.  Cryptographic primitives with potential issues are displayed in yellow; primitives with serious flaws are displayed in red.


.SH POLICY AUDIT
.PP
When the -P/--policy option is used, \fBssh-audit\fP performs a policy audit.  The target's host key types, key exchanges, ciphers, MACs, and other information is compared to a set of expected values defined in the specified policy file.  If everything matches, only a short message stating a passing result is reported.  Otherwise, the field(s) that did not match are reported.

.PP
Policy auditing is helpful for ensuring a group of related servers are properly hardened to an exact specification.

.PP
The set of official built-in policies can be viewed with -L/--list-policies.  Multiple servers can be audited with -T/--targets=<servers.txt>.  Custom policies can be made from an ideal target server with -M/--make-policy=<custom_policy.txt>.


.SH EXAMPLES
.LP
Basic server auditing:
.RS
.nf
ssh-audit localhost
ssh-audit 127.0.0.1
ssh-audit 127.0.0.1:222
ssh-audit ::1
ssh-audit [::1]:222
.fi
.RE

.LP
To run a standard audit against many servers (place targets into servers.txt, one on each line in the format of HOST[:PORT]):
.RS
.nf
ssh-audit -T servers.txt
.fi
.RE

.LP
To audit a client configuration (listens on port 2222 by default; connect using "ssh -p 2222 anything@localhost"):
.RS
.nf
ssh-audit -c
.fi
.RE

.LP
To audit a client configuration, with a listener on port 4567:
.RS
.nf
ssh-audit -c -p 4567
.fi
.RE

.LP
To list all official built-in policies (hint: use their full names with -P/--policy):
.RS
.nf
ssh-audit -L
.fi
.RE

.LP
To run a built-in policy audit against a server (hint: use -L to see list of built-in policies):
.RS
.nf
ssh-audit -P "Hardened Ubuntu Server 20.04 LTS (version 1)" targetserver
.fi
.RE


.LP
To run a custom policy audit against a server (hint: use -M/--make-policy to create a custom policy file):
.RS
.nf
ssh-audit -P path/to/server_policy.txt targetserver
.fi
.RE

.LP
To run a policy audit against a client:
.RS
.nf
ssh-audit -c -P ["policy name" | path/to/client_policy.txt]
.fi
.RE

.LP
To run a policy audit against many servers:
.RS
.nf
ssh-audit -T servers.txt -P ["policy name" | path/to/server_policy.txt]
.fi
.RE

.LP
To create a policy based on a target server (which can be manually edited; see official built-in policies for syntax examples):
.RS
.nf
ssh-audit -M new_policy.txt targetserver
.fi
.RE

.LP
To run a Diffie-Hellman Group Exchange modulus size test using the values 2000 bits, 3000 bits, 4000 bits and 5000 bits:
.RS
.nf
ssh-audit targetserver --gex-test=2000,3000,4000,5000
.fi
.RE

.LP
To run a Diffie-Hellman Group Exchange modulus size test where 2048 bits is the minimum, 3072 bits is the preferred and 5000 bits is the maximum:
.RS
.nf
ssh-audit targetserver --gex-test=2048:3072:5000
.fi
.RE

.LP
To run a Diffie-Hellman Group Exchange modulus size test from 0 bits to 5120 bits in increments of 1024 bits:
.RS
.nf
ssh-audit targetserver --gex-test=0-5120:1024
.fi
.RE

.LP
To run the DHEat DoS attack (monitor the target server's CPU usage to determine the optimal number of concurrent sockets):
.RS
.nf
ssh-audit targetserver --dheat=10
.fi
.RE

.LP
To run the DHEat attack and manually target the diffie-hellman-group-exchange-sha256 algorithm:
.RS
.nf
ssh-audit targetserver --dheat=10:diffie-hellman-group-exchange-sha256
.fi
.RE

.LP
To run the DHEat attack and manually target the diffie-hellman-group-exchange-sha256 algorithm with a very small length of e (resulting in the same effect but without having to send large packets):
.RS
.nf
ssh-audit targetserver --dheat=10:diffie-hellman-group-exchange-sha256:4
.fi
.RE

.LP
To test the number of successful connections per second that can be created with the target using 8 parallel threads (useful for detecting whether connection throttling is implemented by the target):
.RS
.nf
ssh-audit targetserver --conn-rate-test=8
.fi
.RE

.LP
To use 8 parallel threads to create up to 100 connections per second with the target (useful for understanding how much CPU load is caused on the target simply from handling new connections vs excess modular exponentiation when performing the DHEat attack):
.RS
.nf
ssh-audit targetserver --conn-rate-test=8:100
.fi
.RE

.SH RETURN VALUES
When a successful connection is made and all algorithms are rated as "good", \fBssh-audit\fP returns 0.  Other possible return values are:

.RS
.nf
1 = connection error
2 = at least one algorithm warning was found
3 = at least one algorithm failure was found
<any other non-zero value> = unknown error
.fi
.RE

.SH SSH HARDENING GUIDES
Hardening guides for common platforms can be found at: <https://www.ssh\-audit.com/>

.SH BUG REPORTS
Please file bug reports as a Github Issue at: <https://github.com/jtesta/ssh\-audit/issues>

.SH AUTHOR
.LP
\fBssh-audit\fP was originally written by Andris Raugulis <moo@arthepsy.eu>, and maintained from 2015 to 2017.
.br
.LP
Maintainership was assumed and development was resumed in 2017 by Joe Testa <jtesta@positronsecurity.com>.