Man Linux: Main Page and Category List

NAME

       etterfilter  NG-0.7.3  - Filter compiler for ettercap content filtering
       engine

SYNOPSIS

       etterfilter [OPTIONS] FILE

DESCRIPTION

       The etterfilter utility is used to compile  source  filter  files  into
       binary  filter  files that can be interpreted by the JIT interpreter in
       the ettercap(8) filter engine. You have to compile your filter  scripts
       in  order  to  use  them  in  ettercap. All syntax/parse errors will be
       checked at compile time, so you will  be  sure  to  produce  a  correct
       binary filter for ettercap.

       GENERAL OPTIONS

       -o, --output <FILE>
              you  can  specify  the  output file for a source filter file. By
              default the output is filter.ef.

       -t, --test <FILE>
              you can  analyze  a  compiled  filter  file  with  this  option.
              etterfilter  will  print  in  a  human  readable  form  all  the
              instructions contained in it. It is a sort of "disassembler" for
              binary filter files.

       -d, --debug
              prints  some  debug messages during the compilation. Use it more
              than once to increase the debug level ( etterfilter -ddd ...  ).

       -w, --suppress-warnings
              Don’t  exit  on  warnings.  With  this  option the compiler will
              compile the script even if it contains warnings.

       STANDARD OPTIONS

       -v, --version
              Print the version and exit.

       -h, --help
              prints the help screen with a short  summary  of  the  available
              options.

       SCRIPTS SYNTAX
              A   script  is  a  compound  of  instructions.  It  is  executed
              sequentially and you can make branches with the ’if’ statements.
              ’if’  and  ’if/else’ statements are the only supported. No loops
              are implemented. The syntax is almost like C  code  except  that
              you have to put ’if’ blocks into graph parentheses ’{’ ’}’, even
              if they contain only one instruction.

              NOTE: you have to put a space between the ’if’ and the ’(’.  You
              must not put the space between the function name and the ’(’.

              Example:
              if (conditions) { }
              func(args...);

              The  conditions for an ’if’ statement can be either functions or
              comparisons.  Two or more conditions can be linked together with
              logical operators like OR ’||’ and AND ’&&’.

              Example:
              if (tcp.src == 21 && search(DATA.data, "ettercap")) {
              }

              Pay  attention  to  the  operator  precedence.   You  cannot use
              parentheses to group conditions, so be careful with  the  order.
              An  AND  at the beginning of a conditions block will exclude all
              the other tests if it is evaluated  as  false.  The  parsing  is
              left-to-right,  when  an  operator is found: if it is an AND and
              the previous condition is false, all the statement is  evaluated
              as  false;  if  it  is  an  OR  the  parsing goes on even if the
              condition is false.

              Example:
              if (ip.proto == UDP || ip.proto == TCP && tcp.src == 80) {
              }

              if (ip.proto == TCP && tcp.src == 80 || ip.proto == UDP) {
              }

              the former condition will match all udp  or  http  traffic.  The
              latter  is  wrong,  because  if the packet is not tcp, the whole
              condition block will be evaluated as false.  If you want to make
              complex  conditions,  the  best way is to split them into nested
              ’if’ blocks.

              Every instruction in a block must end with a semicolon ’;’.

              Comparisons are implemented with the ’==’ operator  and  can  be
              used  to compare numbers, strings or ip addresses. An ip address
              MUST be enclosed within two single quotes  (eg.  ’192.168.0.7’).
              You  can  also  use the ’less than’ (’<’), ’greater than’ (’>’),
              ’less or equal’ (’<=’) and ’greater or equal’ (’>=’)  operators.
              The lvalue of a comparison must be an offset (see later)

              Example:
              if (DATA.data + 20 == "ettercap" && ip.ttl > 16) {
              }

              Assignments are implemented with the ’=’ operator and the lvalue
              can be an offset (see later). The rvalue can  be  a  string,  an
              integer or a hexadecimal value.

              Example:
              ip.ttl = 0xff;
              DATA.data + 7 = "ettercap NG";

              You  can  also  use the ’inc’ and ’dec’ operations on the packet
              fields. The operators used are ’+=’ and ’-=’. The rvalue can  be
              an integer or a hexadecimal value.

              Example:
              ip.ttl += 5;

       OFFSET DEFINITION
              An offset is identified by a virtual pointer. In short words, an
              offset is a pointer to the packet buffer. The virtual pointer is
              a tuple <L, O, S>, where L is the iso/osi level, O is the offset
              in that level and S is the size of the virtual pointer.  You can
              make algebraic operations on a virtual pointer and the result is
              still an offset. Specifying ’vp  +  n’  will  result  in  a  new
              virtual  pointer  <L,  O+n, S>.  And this is perfectly legal, we
              have changed the internal offset of that level.

              Virtual pointers are  in  the  form  ’name.field.subfield’.  For
              example  ’ip.ttl’  is  the  virtual pointer for the Time To Live
              field in the IP header of a packet. It  will  be  translated  as
              <L=3,  O=9,  S=1>.  Indeed it is the 9th byte of level 3 and its
              size is 1 byte. ’ip.ttl + 1’ is the same as ’ip.proto’ since the
              10th  byte  of the IP header is the protocol encapsulated in the
              IP packet.

              The list of all  supported  virtual  pointers  is  in  the  file
              etterfilter.tbl. You can add your own virtual pointers by adding
              a new table  or  modifying  the  existing  ones.  Refer  to  the
              comments  at  the  beginning  of  the  file  for  the  syntax of
              etterfilter.tbl file.

       SCRIPTS FUNCTIONS

       search(where, what)
              this function searches the string ’what’ in the buffer  ’where’.
              The  buffer  can be either DATA.data or DECODED.data. The former
              is the payload at layer  DATA  (ontop  TCP  or  UDP)  as  it  is
              transmitted   on   the   wire,   the   latter   is  the  payload
              decoded/decrypted by dissectors.
              So, if you want to search in an SSH connection, it is better  to
              use ’DECODED.data’ since ’data’ will be encrypted.
              The string ’what’ can be binary. You have to escape it.

              example:
              search(DATA.data, "\x41\x42\x43")

       regex(where, regex)
              this  function  will  return true if the ’regex’ has matched the
              buffer ’where’.  The  considerations  about  ’DECODED.data’  and
              ’DATA.data’ mentioned for the function ’search’ are the same for
              the regex function.

              NOTE: regex can be used only against a string buffer.

              example:
              regex(DECODED.data, ".*login.*")

       pcre_regex(where, pcre_regex ... )
              this  function  will  evaluate   a   perl   compatible   regular
              expression.  You can match against both DATA and DECODED, but if
              your expression modifies the buffer, it makes sense  to  operate
              only  on  DATA. The function accepts 2 or 3 parameters depending
              on the operation you want. The two parameter form is  used  only
              to match a pattern. The three parameter form means that you want
              to make a substitution. In both cases, the second  parameter  is
              the search string.
              You can use $n in the replacement string. These placeholders are
              referred to the groups  created  in  the  search  string.  (e.g.
              pcre_regex(DATA.data,    "^var1=([:digit:]*)&var2=([:digit:]*)",
              "var1=$2&var2=$1") will swap the value of var1 and var2).
              NOTE: The pcre support is  optional  in  ettercap  and  will  be
              enabled  only  if  you have the libpcre installed.  The compiler
              will warn you if you try to compile a filter that contains  pcre
              expressions  but  you  don’t  have libpcre. Use the -w option to
              suppress the warning.

              example:
              pcre_regex(DATA.data, ".*foo$")
              pcre_regex(DATA.data, "([^ ]*) bar ([^ ]*)", "foo $1 $2")

       replace(what, with)
              this function replaces the string ’what’ with the string ’with’.
              They  can  be binary string and must be escaped. The replacement
              is always performed in DATA.data since is the only payload which
              gets   forwarded.   The   ’DECODED.data’  buffer  is  used  only
              internally and never reaches the wire.

              example:
              replace("ethercap", "ettercap")

       inject(what)
              this function injects the content of the file ’what’  after  the
              packet  being processed. It always injects in DATA.data. You can
              use it to replace the entire packet with a fake  one  using  the
              drop() function right before the inject() command.  In that case
              the filtering engine will drop the current packet and inject the
              fake one.

              example:
              inject("./fake_packet")

       log(what, where)
              this  function  dumps  in the file ’where’ the buffer ’what’. No
              information is stored about the  packet,  only  the  payload  is
              dumped.  So  you will see the stream in the file. If you want to
              log packets in a  more  enhanced  mode,  you  need  to  use  the
              ettercap -L option and analyze it with etterlog(8).
              The  file  ’where’  must  be  writable  to  the user EC_UID (see
              etter.conf(5)).

              example:
              log(DECODED.data, "/tmp/interesting.log")

       msg(message)
              this function displays  a  message  to  the  user  in  the  User
              Messages  window.  It  is  useful to let the user know whether a
              particular filter has been successful or not.

              example:
              msg("Packet filtered successfully")

       drop() this function marks the packet "to be dropped". The packet  will
              not be forwarded to the real destination.

              example:
              drop()

       kill() this function kills the connection that owns the matched packet.
              If it is a TCP connection, a RST is sent to both  sides  of  the
              connection. If it is an UDP connection, an ICMP PORT UNREACHABLE
              is sent to the source of the packet.

              example:
              kill()

       exec(command)
              this function executes a shell command. You have to provide  the
              full  path  to  the  command  since  it  is executed without any
              environment. There is no way to determine  if  the  command  was
              successful  or  not.  Furthermore, it is executed asynchronously
              since it is forked by the main process.

              example:
              exec("/bin/cat /tmp/foo >> /tmp/bar")

       exit() this function causes the filter engine  to  stop  executing  the
              code.  It  is useful to stop the execution of the script on some
              circumstance checked by an ’if’ statement.

              example:
              exit()

EXAMPLES

       Here are some examples of using etterfilter.

       etterfilter filter.ecf -o filter.ef

              Compiles the source filter.ecf into a binary filter.ef

AUTHORS

       Alberto Ornaghi (ALoR) <alor@users.sf.net>
       Marco Valleri (NaGA) <naga@antifork.org>

SEE ALSO

       etter.filter.examples
       ettercap(8)      etterlog(8)      etter.conf(5)      ettercap_curses(8)
       ettercap_plugins(8)