NTOP

The text of filename is copied - ignoring line breaks and comment lines (anything following a #) - into the command line. ntop behaves as if all of the text had simply been typed directly on the command line. For example, if the command line is "-t 3 @d -u ntop" and file d contains just the line ’-d’, then the effective command line is -t 3 -d -u ntop. Multiple @s are permitted. Nested @s (an @ inside the file) are not permitted.

Remember, most ntop options are "sticky", that is they just set an internal flag. Invoking them multiple times doesn’t change ntop’s behavior. However, options that set a value, such as --trace-level, will use the LAST value given: --trace-level 2 --trace-level 3 will run as --trace-level 3.

By default ntop does not maintain a log of HTTP requests to the internal web server. Use this parameter to request logging and to specify the location of the file where these HTTP requests are logged.

Each log entry is in Apache-like style. The only difference between Apache and ntop logs is that an additional column has been added which has the time (in milliseconds) that ntop needed to serve the request. Log entries look like this:

192.168.1.1 - - [04/Sep/2003:20:38:55 -0500] - "GET / HTTP/1.1" 200 1489 4
192.168.1.1 - - [04/Sep/2003:20:38:55 -0500] - "GET /index_top.html HTTP/1.1" 200 1854 4
192.168.1.1 - - [04/Sep/2003:20:38:55 -0500] - "GET /index_inner.html HTTP/1.1" 200 1441 7
192.168.1.1 - - [04/Sep/2003:20:38:56 -0500] - "GET /index_left.html HTTP/1.1" 200 1356 4
192.168.1.1 - - [04/Sep/2003:20:38:56 -0500] - "GET /home_.html HTTP/1.1" 200 154/617 9
192.168.1.1 - - [04/Sep/2003:20:38:56 -0500] - "GET /home.html HTTP/1.1" 200 1100/3195 10
192.168.1.1 - - [04/Sep/2003:20:38:56 -0500] - "GET /About.html HTTP/1.1" 200 2010 10

Although this parameter is called a ’path’, it is actually the complete file name of the access log.

This parameter disables protocol decoders.

Protocol decoders examine and collect information about layer 2 protocols such as NetBIOS or Netware SAP, as well as about specific tcp/ip (layer 3) protocols, such as DNS, http and ftp.

This support is specifically coded for each protocol and is different from the capability to count raw information (packets and bytes) by protocol specified by the -p | --protocols parameter, below.

Decoding protocols is a significant consumer of resources. If the ntop host is underpowered or monitoring a very busy network, you may wish to disable protocol decoding via this parameter. It may also be appropriate to use this parameter if you believe that ntop has problems handling some protocols that occur on your network.

Even if decoding is disabled, ftp-data traffic is still decoded to look for passive ftp port commands.

Use this parameter to prevent idle hosts from being purged from memory.

By default idle hosts are periodically purged from memory. An idle host is identified when no packets from or to that host have been monitored for the period of time defined by the value of PARM_HOST_PURGE_MINIMUM_IDLE in globals-defines.h.

If you use this option, all hosts - active and idle - are retained in memory for the duration of the ntop run.

P2P users, port scans, popular web servers and other activity will cause ntop to record data about a large number of hosts. On an active network, this will consume a significant - and always growing - amount of memory. It is strongly recommended that you use a filtering expression to limit the hosts which are stored if you use --sticky-hosts.

The idle purge is a statistical one - a random selection of the eligible hosts will be purged during each cycle. Thus it is possible on a busy system for an idle host to remain in the ntop tables and appear ’active’ for some considerable time after it is truly idle.

This parameter causes ntop to become a daemon, i.e. a task which runs in the background without connection to a specific terminal. To use ntop other than as a casual monitoring tool, you probably will want to use this option.

WARNING: If you are running as a daemon, the messages from ntop will be ’printed’ on to stdout and thus dropped. You probably don’t want to do this. So remember to also use the -L or --use-syslog options to save the messages into the system log.

This defines the maximum number of lines that ntop will display on each generated HTML page. If there are more lines to be displayed than this setting permits, only part of the data will be displayed. There will be page forward/back arrows placed at the bottom of the page for navigation between pages.

By default, ntop captures traffic from network interface cards (NICs) or from netFlow/sFlow probes. However, ntop can also read data from a file - typically a tcpdump capture or the output from one of the ntop packet capture options.

if you specify -f, ntop will not capture any traffic from NICs during or after the file has been read. netFlow/sFlow capture - if enabled - would still be active.

This option is mostly used for debug purposes.

By default, ntop tracks all hosts that it sees from packets captured on the various NICs. Use this parameter to tell ntop to capture data only about local hosts. Local hosts are defined based on the addresses of the NICs and those networks identified as local via the -m | --local-subnets parameter.

This parameter is useful on large networks or those that see many hosts, (e.g. a border router or gateway), where information about remote hosts is not desired/required to be tracked.

Print help information for ntop, including usage and parameters.

Specifies the network interface or interfaces to be used by ntop for network monitoring.

If multiple interfaces are used (this feature is available only if ntop is compiled with thread support) their names must be separated with a comma. For instance -i "eth0,lo".

If not specified, the default is the first Ethernet device, e.g. eth0. The specific device that is ’first’ is highly system dependent. Especially on systems where the device name reflects the driver name instead of the type of interface.

By default, traffic information obtained by all the interfaces is merged together as if the traffic was seen by only one interface. Use the -M parameter to keep traffic separate by interface.

Under Windows, the parameter value is either the number of the interface or its name, e.g. {6252C14C-44C9-49D9-BF59-B2DC18C7B811}. Run ntop -h to see a list of interface name-number mappings (at the end of the help information).

When this parameter is used, the current filter expression is displayed in an extra frame and thus is always visible. This extra frame contains other information, including the report creation date, ntop version information and the active interfaces.

This parameter causes a dump file to be created of the network traffic captured by ntop in tcpdump (pcap) format. This file is useful for debug, and may be read back into ntop by the -f | --traffic-dump-file parameter. The dump is made after processing any filter expression ( never even sees filtered packets).

The output file will be named <path>/<log>.<device>.pcap (Windows: <path>/<log>.pcap ), where <path> is defined by the -O | --output-packet-path parameter and <log> is defined by this -l | --pcap-log parameter.

ntop determines the ip addresses and netmasks for each active interface. Any traffic on those networks is considered local. This parameter allows the user to define additional networks and subnetworks whose traffic is also considered local in ntop reports. All other hosts are considered remote.

Commas separate multiple network values. Both netmask and CIDR notation may be used, even mixed together, for instance "131.114.21.0/24,10.0.0.0/255.0.0.0".

The local subnet - as defined by the interface address(es) - is/are always local and do not need to be specified. If you do give the same value as a NIC’s local address, a harmless warning message is issued.

By default, ntop resolves IP addresses using a combination of active (explicit) DNS queries and passive sniffing. Sniffing of DNS responses occurs when ntop receives a network packet containing the response to some other user’s DNS query. ntop captures this information and enters it into ntop’s DNS cache, in expectation of shortly seeing traffic addressed to that host. This way ntop significantly reduces the number of DNS queries it makes.

This parameter causes ntop to skip DNS resolution, showing only numeric IP addresses instead of the symbolic names. This option can useful when the DNS is not present or quite slow.

ntop is a hybrid layer 2/3 network monitor. That is, it uses both the lower level, physical device address - the MAC (Media Access Control) address - and the higher level, logical, tcp/ip address (the familiar www.ntop.org or 131.114.21.9 address). This allows ntop to link the logical addresses to a physical machine with multiple addresses (This occurs with virtual hosts or additional addresses assigned to the interface, etc.) to present consolidated reporting.

This parameter specifies that ntop should not trust the MAC addresses but just use the IP addresses.

Normally, since the MAC address must be globally unique, the dual nature of ntop is a benefit and provides far better information about the network than is available via a pure layer 2 or pure layer 3 monitor.

Under certain circumstances - whenever ntop is started on an interface where MAC addresses cannot be really trusted - you may require this option.

Situations which may require this option include port/VLAN mirror, some cases with switches and spanning tree protocol, and (reportedly) some specific models of Ethernet switches which re-write MAC addresses of the packets they process. Normally, you discover that this option is necessary when you observe that hosts seem to change their addresses or information about different machines get lumped together.

Note that with this option, information which is dependent upon the MAC addresses (non tcp/ip protocols like IPX) will not be collected nor displayed.

This parameter is used to specify the TCP/UDP protocols that ntop will monitor. The format is <label>=<protocol list> [, <label>=<protocol list>], where label is used to symbolically identify the <protocol list>. The format of <protocol list> is <protocol>[|<protocol>], where <protocol> is either a valid protocol specified inside the /etc/services file or a numeric port range (e.g. 80, or 6000-6500).

A simple example is --protocols="HTTP=http|www|https|3128,FTP=ftp|ftp-data", which reduces the protocols displayed on the "IP" pages to three:

Host                      Domain Data          HTTP   FTP   Other IP
ns2.attbi.com             <flag>  954 63.9 %      0     0        954
64.124.83.112.akamai.com  <flag>  240 16.1 %    240     0          0
64.124.83.99.akamai.com   <flag>  240 16.1 %    240     0          0
toolbarqueries.google.com <flag>   60 4.0 %      60     0          0

If the <protocol list> is very long you may store it in a file (for instance protocol.list). To do so, specify the file name instead of the <protocol list> on the command line. e.g. ntop -p protocol.list

If the -p parameter is omitted the following default value is used:

  FTP=ftp|ftp-data
  HTTP=http|www|https|3128     3128 is Squid, the HTTP cache
  DNS=name|domain
  Telnet=telnet|login
  NBios-IP=netbios-ns|netbios-dgm|netbios-ssn
  Mail=pop-2|pop-3|pop3|kpop|smtp|imap|imap2
  DHCP-BOOTP=67-68
  SNMP=snmp|snmp-trap
  NNTP=nntp
  NFS=mount|pcnfs|bwnfs|nfsd|nfsd-status
  X11=6000-6010
  SSH=22

 Peer-to-Peer Protocols
  ----------------------
  Gnutella=6346|6347|6348
  Kazaa=1214
  WinMX=6699|7730
  DirectConnect=0      Dummy port as this is a pure P2P protocol
  eDonkey=4661-4665

 Instant Messenger
  -----------------
  Messenger=1863|5000|5001|5190-5193

NOTE: To resolve protocol names to port numbers, they must be specified in the system file used to list tcp/udp protocols and ports, which is typically /etc/services file. You will have to match the names in that file, exactly. Missing or unspecified (non-standard) ports must be specified by number, such as 3128 in our examples above.

If you have a file named /etc/protocols, don’t get confused by it, as that’s the Ethernet protocol numbers, which are not what you’re looking for.

This parameter tells ntop to create a dump file of suspicious packets.

There are many, many, things that cause a packet to be labeled as ’suspicious’, including:

  Detected ICMP fragment
  Detected Land Attack against host
  Detected overlapping/tiny packet fragment
  Detected traffic on a diagnostic port
  Host performed ACK/FIN/NULL scan
  Host rejected TCP session
  HTTP/FTP/SMTP/SSH detected at wrong port
  Malformed TCP/UDP/ICMP packet (packet too short)
  Packet # %u too long
  Received a ICMP protocol Unreachable from host
  Sent ICMP Administratively Prohibited packet to host
  Smurf packet detected for host
  TCP connection with no data exchanged
  TCP session reset without completing 3-way handshake
  Two MAC addresses found for the same IP address
  UDP data to a closed port
  Unknown protocol (no HTTP/FTP/SMTP/SSH) detected (on port 80/21/25/22)
  Unusual ICMP options

When this parameter is used, one file is created for each network interface where suspicious packets are found. The file is in tcpdump (pcap) format and is named <path>/ntop-suspicious-pkts.<device>.pcap, where <path> is defined by the -O | --output-packet-path parameter.

Specifies the delay (in seconds) between automatic screen updates for those generated HTML pages which support them. This parameter allows you to leave your browser window open and have it always displaying nearly real-time data from ntop.

The default is 3 seconds. Please note that if the delay is very short (1 second for instance), ntop might not be able to process all of the network traffic.

Use this parameter to prevent from setting the interface(s) into promiscuous mode.

An interface in promiscuous mode will accept ALL Ethernet frames, regardless of whether they directed (addressed) to the specific network interface (NIC) or not. This is an essential part of enabling ntop to monitor an entire network. (Without promiscuous mode, ntop will only see traffic directed to the specific host it is running on, plus broadcast traffic such as the arp and dhcp protocols.

Even if you use this parameter, the interface could well be in promiscuous mode if another application enabled it.

ntop passes this setting on to libpcap, the packet capture library. On many systems, a non-promiscuous open of the network interface will fail, since the libpcap function on most systems require it to capture raw packets ( ntop captures raw packets so that we may view and analyze the layer 2 - MAC - information).

Thus on most systems, ntop must probably still be started as root, and this option is largely ornamental. If it fails, you will see a ***FATALERROR*** message referring to pcap_open_live() and then an information message, "Sorry, but on this system, even with -s, it appears that ntop must be started as root".

This parameter specifies the ’information’ level of messages that you wish ntop to display (on stdout or to the log). The higher the trace level number the more information that is displayed. The trace level ranges between 0 (no trace) and 5 (full debug tracings).

The default trace value is 3.

Trace level 0 is not quite zero messages. Fatal errors and certain startup/shutdown messages are always displayed. Trace level 1 is used to display errors only, level 2 for both errors and warnings, and level 3 displays error, warning and informational messages.

Trace level 4 is called ’noisy’ and it is - generating many messages about the internal functioning of ntop. Trace level 5 is ’noisy’ plus --log-extra 1, i.e. all possible messages, with a file:line tag prepended to every message.

Specifies the user ntop should run as after it initializes.

ntop must normally be started as root so that it has sufficient privileges to open the network interfaces in promiscuous mode and to receive raw frames. See the discussion of -s | --no-promiscuous above, if you wish to try starting ntop as a non-root user.

Shortly after starting up, ntop becomes the user you specify here, which normally has substantially reduced privileges, such as no login shell. This is the userid which owns ntop’s database and output files.

The value specified may be either a username or a numeric user id. The group id used will be the primary group of the user specified.

If this parameter is not specified, ntop will try to switch first to ’nobody’ and then to ’anonymous’ before giving up.

NOTE: This should not be root unless you really understand the security risks. In order to prevent this by accident, the only way to run ntop as root is to explicitly specify -u root. Don’t do it.

ntop offers an embedded web server to present the information that has been so painstakingly gathered. An external HTTP server is NOT required NOR supported. The ntop web server is embedded into the application. These parameters specify the port (and optionally the address (i.e. interface)) of the ntop web server.

For example, if started with -w 3000 (the default port), the URL to access ntop is http://hostname:3000/. If started with a full specification, e.g. -w 192.168.1.1:3000, ntop listens on only that address/port combination.

If -w is set to 0 the web server will not listen for http:// connections.

-W operates similarly, but controls the port for the https:// connections.

Some examples:

ntop -w 3000 -W 0 (this is the default setting) HTTP requests on port 3000 and no HTTPS.

ntop -w 80 -W 443 Both HTTP and HTTPS have been enabled on their most common ports.

ntop -w 0 -W 443 HTTP disabled, HTTPS enabled on the common port.

Certain sensitive, configuration pages of the ntop web server are protected by a userid/password. By default, these are the user/URL administration, filter, shutdown and reset stats are password protected and are accessible initially only to user admin with a password set during the first run of ntop.

Users can modify/add/delete users/URLs using ntop itself - see the Admin tab.

The passwords, userids and URLs to protect with passwords are stored in a database file. Passwords are stored in an encrypted form in the database for further security. Best practices call for securing that database so that only the ntop user can read it.

There is a discussion in docs/FAQ about further securing the ntop environment.

This parameter disables TCP session tracking. Use it for better performance or when you don’t really need/care to track sessions.

This parameter is used to start ntop , set the admin password and quit. It is quite useful for installers that need to automatically set the password for the admin user.

-A and --set-admin-password (without a value) will prompt the user for the password.

You may also use this parameter to set a specific value using --set-admin-password=value. The = is REQUIRED and no spaces are permitted!

If you attempt to run ntop as a daemon without setting a password, a FATAL ERROR message is generated and ntop stops.

Filters allows the user to restrict the traffic seen by ntop on just about any imaginable item.

The filter expression is set at run time by this parameter, but it may be changed during the ntop run on the Admin | Change Filter web page.

The basic format is -B filter , where the quotes are REQUIRED

The syntax of the filter expression uses the same BPF (Berkeley Packet Filter) expressions used by other packages such as tcpdump

For instance, suppose you are interested only in the traffic generated/received by the host jake.unipi.it. ntop can then be started with the following filter:

ntop -B src host jake.unipi.it or dst host jake.unipi.it

or in shorthand:

ntop -B host jake.unipi.it or host jake.unipi.it

See the ’expression’ section of the tcpdump man page - usually available at http://www.tcpdump.org/tcpdump_man.html - for further information and the best quick guide to BPF filters currently available.

This identifies the local domain suffix, e.g. ntop.org. It may be necessary, if ntop is having difficulty determining it from the interface.

It is used to specify network flows similar to more powerful applications such as NeTraMet. A flow is a stream of captured packets that match a specified rule. The format is

" <flow-label>=’<matching expression>’[,<flow-label>=’<matching expression>’] "

, where the label is used to symbolically identify the flow specified by the expression. The expression format is specified in the appendix. If an expression is specified, then the information concerning flows can be accessed following the HTML link named ’List NetFlows’.

For instance define two flows with the following expression " LucaHosts=’host jake.unipi.it or host pisanino.unipi.it’,GatewayRoutedPkts=’gateway gateway.unipi.it’ ".

All the traffic sent/received by hosts jake.unipi.it or pisanino.unipi.it is collected by ntop and added to the LucaHosts flow, whereas all the packet routed by the gateway gateway.unipi.it are added to the GatewayRoutedPkts flow. If the flows list is very long you may store in a file (for instance flows.list) and specify the file name instead of the actual flows list (in the above example, this would be ’ntop -F flows.list’).

Note that the double quotations around the entire flow expression are required.

Use this parameter to simplify application debug. It does three things: 1. Does not fork() on the "read only" html pages. 2. Displays mutex values on the configuration (info.html) page. 3. (If available - glibc/gcc) Activates an automated backtrace on application errors.

Use this parameter to send log messages to the system log instead of stdout.

-L and the simple form --use-syslog use the default log facility, defined as LOG_DAEMON in the #define symbol DEFAULT_SYSLOG_FACILITY in globals-defines.h.

The complex form, --use-syslog=facility will set the log facility to whatever value (e.g. local3, security) you specify. The = is REQUIRED and no spaces are allowed!

This setting applies both to ntop and to any child fork()ed for reporting. If this parameter is not specified, any fork()ed child will use the default value and will log it’s messages to the system log (this occurs because the fork()ed child must give up it’s access to the parents stdout).

Because various systems do not make the permissible names available, we have a table at the end of globals-core.c. Look for myFacilityNames.

By default, ntop merges the data collected from all of the interfaces (NICs) it is monitoring into a single set of counters.

If you have a simple network, say a small LAN with a connection to the internet, merging data is good as it gives you a better picture of the whole network. For larger, more complex networks, this may not be desirable. You may also have other reasons for wishing to monitor each interface separately, for example DMZ vs. LAN traffic.

This option instructs ntop not to merge network interfaces together. This means that ntop will collect statistics for each interface and report them separately.

Only ONE interface may be reported on at a time - use the Admin | Switch NIC option on the web server to select which interface to report upon.

Note that activating either the netFlow and/or sFlow plugins will force the setting of -M. Once enabled, you cannot go back.

This parameter defines the base path for the ntop-suspicious-pkts.XXX.pcap and normal packet dump files.

If this parameter is not specified, the default value is the config.h parameter CFG_DBFILE_DIR, which is set during ./configure from the --localstatedir= parameter. If --localstatedir is not specified, it defaults to the --prefix value plus /var (e.g. /usr/local/var).

Be aware that this may not be what you expect when running ntop as a daemon or Windows service. Setting an explicit and absolute path value is STRONGLY recommended if you use this facility.

These parameters specify where ntop stores database files.

There are two types, ’temporary’ - that is ones which need not be retained from ntop run to ntop run, and ’permanent’, which must be retained (or recreated).

The ’permanent’ databases are the preferences, "prefsCache.db" and the password file, "ntop_pw.db". These are stored in the -P | --db-file-path specified location.

Certain plugins use the -P | --db-file-path specified location for their database ("LsWatch.db") or (as a default value) for files (.../rrd/...).

The ’temporary’ databases are the address queue, "addressQueue.db", the cached DNS resolutions, "dnsCache.db" and the MAC prefix (vendor table), "macPrefix.db".

If only -P | --db-file-path is specified, it is used for both types of databases.

The directories named must allow read/write and file creation by the ntop user. For security, nobody else should have even read access to these files.

Note that the default value is the config.h parameter CFG_DBFILE_DIR. This is set during ./configure from the --localstatedir= parameter. If --localstatedir is not specified, it defaults to the --prefix value plus /var (e.g. /usr/local/var).

This may not be what you expect when running ntop as a daemon or Windows service.

Note that on versions of ntop prior to 2.3, these parameters defaulted to "." (the current working directory, e.g. the value returned by the pwd command) and caused havoc as it was different when ntop was run from the command line, vs. run via cron, vs. run from an initialization script.

Setting an explicit and absolute path value is STRONGLY recommended.

Specifies the URL of the mapper.pl utility.

If provided, ntop creates a clickable hyperlink on the ’Info about host xxxxxx’ page to this URL by appending ?host=xxxxx. Any type of host lookup could be performed, but this is intended to lookup the geographical location of the host.

A cgi-based mapper interface to http://www.multimap.com is part of the ntop distribution [see www/Perl/mapper.pl]).

Prints ntop version information and then exits.

(See the joint documentation with the -w parameter, above)

Return ntop to the old (v2.1) behavior on a memory error. The default of stopcap enabled makes the web interface available albeit with static content until ntop is shutdown.

This optional parameter controls the addition of more information to each log message. Both choices are useful in different ways, for debugging and for using log watching and filtering packages.

Setting 1 adds a [file:line] to the beginning of the log message. Setting 2 adds a [MSGIDnnnnnnn] tag at the end of the log message. The nnnnnnn value should be unique number for every message and should be stable across ntop releases.

ntop sets completed sessions as ’timed out’ and then purge them almost instantly, which is not the behavior you might expect from the discussions about purge timeouts. This switch makes ntop respect the timeouts for completed sessions. It is NOT the default because a busy web server may have 100s or 1000s of completed sessions and this would significantly increase the amount of memory ntop uses.

ntop uses sched_yield() calls for better interactive performance. Under some situations, primarily under RedHat Linux 8.0, this can deadlock, causing the ntop web server to stop responding, although ntop appears to still be operational according to the ps command. Use this switch to disable these calls, IF you are seeing deadlocks.