docs/diploma

annotate thesis/tex/5-Improvements.tex @ 327:d13bddf0f994

added a label
author meillo@marmaro.de
date Fri, 23 Jan 2009 12:58:52 +0100
parents 8671d9c0f29a
children 2b1da14922f7
rev   line source
meillo@246 1 \chapter{Improvement plans}
meillo@89 2
meillo@317 3 The last chapter came to the result that further development is best done in a double-strategy. First the existing code base should be improved to satisfy the most important needs in order to make it usable for some more time. Then \masqmail\ should get redesigned from scratch and rebuild to gain a secure and modern \MTA\ architecture for the future.
meillo@109 4
meillo@317 5 This chapter finally gives concrete suggestions \emph{how} to realize these plans.
meillo@249 6
meillo@317 7 The first part covers the short-time goals which base on current code. The second part deals with the long-time goal---the redesign.
meillo@89 8
meillo@184 9
meillo@184 10
meillo@184 11
meillo@287 12 \section{Based on current code}
meillo@125 13
meillo@249 14 The first three \TODO{}s are implementable by improving the current code or by adding wrappers or interposition filters. The following sections describe solution approaches to do that work.
meillo@184 15
meillo@184 16
meillo@184 17
meillo@326 18 \subsection{Encryption}
meillo@246 19
meillo@317 20 Encryption should be the first funtionality to add to the current code. This requirement was already discussed on page \pageref{requirement-encryption}. As explained there, \NAME{STARTTLS} encryption---as defined in \RFC\,2487---should be added to \masqmail.
meillo@246 21
meillo@317 22 Adding encryption requires changes mainly in three source files: \path{smtp_in.c}, \path{smtp_out.c}, and in \path{conf.c}.
meillo@246 23
meillo@317 24 The first file includes the functionality for the \SMTP\ server. It needs to offer \NAME{STARTTLS} support to clients and needs to initiate the encryption when the client requests it. Additionally, the server should be able to insist on encryption before it accepts any message.
meillo@246 25
meillo@317 26 The second file includes the functionality for the \SMTP\ client. It should start the encryption by issuing the \NAME{STARTTLS} keyword if the server supports it. It should be possible to send messages only if encryption is possible.
meillo@246 27
meillo@317 28 The third file controls the configuration files. New configuration option need to be added. The encryption policy for incoming connections needs to be defined. Three choises seem necessary: no encryption, offer encryption, insist on encryption. The encryption policy for outgoing connections should be part of each route setup. The options are the same: never encrypt, encrypt if possible, insist on encryption.
meillo@246 29
meillo@317 30 \NAME{STARTTLS} uses \NAME{TLS} encryption which is based on certificates. Thus the \MTA\ needs its own certificate. This should be generated during installation. A third party application like \name{openssl} should be taken for this job. The encryption itself should also be done using an available library. Open\NAME{SSL} or a substitute like Gnu\NAME{TLS} does then become a dependency for \masqmail. Gnu\NAME{TLS} seems to be the better choice because the Open\NAME{SSL} license is incompatible to the \NAME{GPL}, under which \masqmail\ and Gnu\NAME{TLS} are covered.
meillo@246 31
meillo@317 32 User definable paths to \masqmail's secret key, \masqmail's certificate, and the public certificates of trusted \name{Certificate Authorities} (short: \NAME{CA}s) are also nice to have.
meillo@246 33
meillo@246 34
meillo@317 35 << TLS patch of qmail >>
meillo@246 36
meillo@246 37 %postfix: main.cf
meillo@317 38 %
meillo@246 39 % smtpd_use_tls = yes
meillo@246 40 % smtpd_tls_received_header = no (does not log in received headers)
meillo@246 41 %
meillo@246 42 % smtpd_tls_key_file = /etc/postfix/key.pem
meillo@246 43 % smtpd_tls_cert_file = /etc/postfix/cert.pem
meillo@246 44 % smtpd_tls_CA_file = /etc/postfix/CAcert.pem
meillo@246 45 %
meillo@246 46 % smtp_use_tls = yes (use TLS for sending)
meillo@246 47 % smtp_tls_key_file = /etc/postfix/key.pem
meillo@246 48 % smtp_tls_cert_file = /etc/postfix/cert.pem
meillo@246 49 % smtp_tls_CA_file = /etc/postfix/CAcert.pem
meillo@246 50
meillo@246 51
meillo@246 52
meillo@184 53
meillo@326 54 \subsection{Authentication}
meillo@125 55
meillo@317 56 Authentication is the second function to add; it is important to restrict the access to \masqmail, especially for mail relay. The requirements for authentication where identified on page \pageref{requirement-authentication}.
meillo@184 57
meillo@317 58 Static access restriction, based on the \NAME{IP} address is already possible by using \name{TCP Wrappers}. This makes it easy to refuse all connections from outside the local net for example, which is a good prevention of being an open relay. More detailed static restrictions, like splitting between mail for the system and mail to relay, should not be added to the current code. This may be a concern for the new design.
meillo@277 59
meillo@317 60 Of the dynamic, secret based, authentication methods (\SMTP-after-\NAME{POP}, \SMTP\ authentication, and certificates) the first one drops out as it requires a \NAME{POP} server running on the same or a trusted host. \NAME{POP} servers are rare on workstations and home servers do also not regularly include them. Thus it is no option for \masqmail.
meillo@277 61
meillo@317 62 Authentication based on certificates does suffer from the certificate infrastructure that is required. Although certificates are already used for encryption, its management overhead prevented wide spread usage for authentication.
meillo@184 63
meillo@317 64 \SMTP\ authentication (also refered to as \NAME{SMTP-AUTH}) support is easiest received by using a \name{Simple Authentication and Security Layer} (short: \NAME{SASL}) implementation. \person{Dent} sees in \NAME{SASL} the best solution for dynamic authentication of users:
meillo@184 65 \begin{quote}
meillo@184 66 %None of these add-ons is an ideal solution. They require additional code compiled into your existing daemons that may then require special write accesss to system files. They also require additional work for busy system administrators. If you cannot use any of the nonauthenticating alternatives mentioned earlier, or your business requirements demand that all of your users' mail pass through your system no matter where they are on the Internet, SASL is probably the solution that offers the most reliable and scalable method to authenticate users.
meillo@184 67 None of these [authentication methods] is an ideal solution. They require additional code compiled into your existing daemons that may then require special write accesss to system files. They also require additional work for busy system administrators. If you cannot use any of the nonauthenticating alternatives mentioned earlier, or your business requirements demand that all of your users' mail pass through your system no matter where they are on the Internet, \NAME{SASL} is probably the solution that offers the most reliable and scalable method to authenticate users.
meillo@218 68 \hfill\cite[page 44]{dent04}
meillo@184 69 \end{quote}
meillo@184 70
meillo@324 71 These days is \NAME{SMTP-AUTH}, which is defined in \RFC\,2554, supported by most email clients. If encryption is used then even insecure authentication methods like \NAME{PLAIN} and \NAME{LOGIN} become secure.
meillo@277 72
meillo@324 73 \masqmail\ best uses an available \NAME{SASL} library. \name{Cyrus} \NAME{SASL} is used by \postfix\ and \sendmail. It is a complete framework that makes use of existing authentication concepts like \path{/etc/passwd} or \NAME{PAM}. As advantage it can be included in existing user data bases. \name{gsasl} is an alternative. It comes as a library which helps on deciding for a method and on generating the appropriate dialog data; the actual transmission of the data and the authentication against some database is left open to the programmer. \name{gsasl} is used by \name{msmtp} for example. It seems best to give both concepts a try and decide then which one to use.
meillo@317 74
meillo@324 75 Currently, outgoing connections already feature \SMTP-\NAME{AUTH} but only in a hand-coded way. It is to decide wether it remains as it is or gets replaced by the \NAME{SASL} approach, that is used for incoming connections. The decision should be based on the estimated time until the new design is usable.
meillo@324 76
meillo@324 77 Authentication needs code changes at the same places as encryption. The relevant code files are \path{smtp_in.c}, \path{smtp_out.c}, and \path{conf.c}.
meillo@324 78
meillo@324 79 The server code, to authenticate clients, must be added to \path{smtp_in.c} and the configuration options to \path{conf.c}. Several configuration options should be provided: the authentication policy (no authentication, offer authentication, insist on authentication), the authentication backend (if several are supported), an option to refuse plain text methods (\NAME{PLAIN} and \NAME{LOGIN}), and one to require encryption before authentication.
meillo@324 80
meillo@324 81 If the authentication code for outgoing connects shall be changed too, it must be done in \path{smtp_out.c}. The configuration options are already present.
meillo@324 82
meillo@324 83
meillo@324 84 About the authentication backend. For a small \MTA\ like \masqmail, it seems preferrable to store the login data in a text file under \masqmail's control. This is the most simple choice for many usage scenarios. But using a central authentication facility has advantages in larger setups too. \name{Cyrus} \NAME{SASL} supports both, so there is no problem. If \name{gsasl} is chosen, it seems best to start with an authentication file under \masqmail's control.
meillo@324 85
meillo@324 86
meillo@324 87
meillo@317 88
meillo@317 89
meillo@317 90 << Compare static with dynamic authentication: pros and cons; usecases: when to use what; >>
meillo@317 91
meillo@317 92 << how could this be covered by architecture (e.g. smtp submission). >>
meillo@184 93
meillo@184 94
meillo@184 95
meillo@324 96
meillo@324 97
meillo@326 98 \subsection{Security}
meillo@184 99
meillo@324 100 Improvements to \masqmail's security are an important requirement and are the third task to work on. Retrofitting security \emph{into} \masqmail\ is not or hardly possible as it was explained in section \ref{sec:discussion-further-devel}. But adding wrappers and interposition filters can be a large step towards security.
meillo@324 101
meillo@324 102 At first mail security layers like \name{smap} come to mind. The market share analysis in section \ref{sec:market-share} identified such software. This is an interposition filter that stands between the untrusted network and the \MTA. It accepts mail in replacement for the \MTA\ (also called \name{proxy}) in order to separate the \MTA\ from the untrusted network.
meillo@324 103
meillo@324 104 The work \name{smap} does is described in \cite{cabral01}: \name{smap} accepts messages as proxy for the \MTA\ and puts it into a queue. \name{smapd} a brother program runs as daemon and watches for new messages in the queue which it submits into the \MTA\ then.
meillo@324 105
meillo@324 106 Because the \MTA\ does not listen for connections from extern now, it is not directly attackable. But the \MTA\ can not react on relaying and spam on itself anymore because it has no direct connection to the mail sender. This job needs to be covered by the proxy now. Similar is the situation for encryption and authentication. However, care must be taken that the proxy stays small and simple as its own security will suffer otherwise.
meillo@324 107
meillo@324 108 The advantage is that the \MTA\ itself needs not to bother much with untrusted environments. And a small proxy cares only about that work.
meillo@324 109
meillo@324 110 \name{smap} is non-free software and thus no general choice for \masqmail. A way to achieve a similar setup would be to copy \masqmail\ and strip one copy to the bare minimum what is needed for the proxy job. \name{setuid} could be removed and root privilege too if \name{inetd} is used. This hardens the proxy instance.
meillo@324 111
meillo@326 112 Mail from extern would then come through the proxy into the system. Mail from the local host and from the local network could be directly accepted by the normal \masqmail, if those locations are considered trusted. But it seems better to have them use the proxy too, or maybe a second proxy instance with different policy.
meillo@324 113
meillo@326 114 The here described setup comes close to the structure of the incoming channels in the new design which is described in \ref{sec:new-design}. This shows the possibilities of the here chosen approach. %fixme: rethink this sentence
meillo@184 115
meillo@184 116
meillo@326 117 \subsubsection*{A concrete setup}
meillo@184 118
meillo@326 119 A stripped down proxy needs to be created. It should only be able to receive mail via \SMTP, encrypt the communication, authenticate clients, and send mail out via \SMTP\ to an internal socket (named ``X'' in the figure). This is a straight forward task. The normal \masqmail\ instance runs on the system too. It takes input from \name{stdin} (by calling the \path{sendmail} command) and via \SMTP\ where it listens on an internal socket (named ``X'' in the figure). Outgoing mail is handled without difference to a regular setup. Figure \ref{fig:proxy-setup} depicts the setup.
meillo@317 120
meillo@326 121 \begin{figure}
meillo@326 122 \begin{center}
meillo@326 123 \includegraphics[scale=0.75]{img/proxy-setup.eps}
meillo@326 124 \end{center}
meillo@326 125 \caption{A setup with a proxy}
meillo@326 126 \label{fig:proxy-setup}
meillo@326 127 \end{figure}
meillo@288 128
meillo@246 129
meillo@298 130 \subsubsection*{Spam and malware handling}
meillo@277 131
meillo@326 132 The presented setup is the same as the one with two \MTA\ instances and a scanner application in between, which was suggested to add spam and malware scanner afterwards to an \MTA. This is a fortunate conincidence, because a scanner like \name{amavis} can simply be put in replace for the internal socket ``X''.
meillo@277 133
meillo@277 134
meillo@246 135
meillo@246 136
meillo@326 137 \subsubsection*{Conditional compilation}
meillo@326 138 << conditional compilation >>
meillo@326 139
meillo@326 140
meillo@326 141
meillo@246 142
meillo@246 143
meillo@246 144
meillo@246 145
meillo@246 146
meillo@246 147
meillo@246 148
meillo@246 149
meillo@246 150
meillo@246 151
meillo@246 152
meillo@246 153
meillo@246 154
meillo@246 155
meillo@285 156 \section{A new design}
meillo@326 157 \label{sec:new-design}
meillo@246 158
meillo@249 159 The last chapter identified the requirements for a modern and securt \masqmail. Now the various jobs of an \MTA\ get assigned to modules, of which the new architecture is created. It is inspired by existing \MTA{}s and driven by the identified requirements.
meillo@246 160
meillo@249 161 One wise experience was kept in mind during the design: ``Many times in life, getting off to the right start makes all the difference.'' \cite[page~32]{graff03}.
meillo@246 162
meillo@246 163
meillo@326 164 Major design ideas of the design were:
meillo@246 165 \begin{itemize}
meillo@246 166 \item free the internal system from in and out channels
meillo@246 167 \item arbitrary protocol handlers have to be addable afterwards
meillo@246 168 \item a single facility for scanning (all mail goes through it)
meillo@246 169 \item concentrate on mail transfer
meillo@246 170 \end{itemize}
meillo@246 171
meillo@246 172
meillo@326 173
meillo@326 174 \subsection{Architectural design}
meillo@326 175
meillo@246 176 \subsubsection*{Incoming channels}
meillo@246 177
meillo@246 178 \sendmail-compatible \mta{}s must support at least two incoming channels: mail submitted using the \sendmail\ command, and mail received via the \SMTP\ daemon. It is therefor common to split the incoming channel into local and remote. This is done by \qmail\ and \postfix. The same way is \person{Hafiz}'s view.
meillo@246 179
meillo@246 180 In contrast is \name{sendmail X}: Its locally submitted messages go to the \SMTP\ daemon, which is the only connection towards the mail queue. %fixme: is it a smtp dialog? or a second door?
meillo@248 181 \person{Finch} proposes a similar approach. He wants the \texttt{sendmail} command to be a simple \SMTP\ client that contacts the \SMTP\ daemon of the \MTA\ like it is done by connections from remote. The advantage here is one single module where all \SMTP\ dialog with submitters is done. Hence one single point to accept or refuse incoming mail. Additionally does the module to put mail into the queue not need to be \name{setuid} or \name{setgid} because it is only invoked from the \SMTP\ daemon. The \MTA's architecture would become simpler and common tasks are not duplicated in modules that do similar jobs.
meillo@246 182
meillo@246 183 But merging the input channels in the \SMTP\ daemon makes the \MTA\ heavily dependent on \SMTP\ being the main mail transfer protocol. To \qmail\ and \postfix\ new modules to support other ways of message receival may be added without change of other parts of the system. Also is it better to have more independent modules if each one is simpler then.
meillo@246 184
meillo@246 185 With the increasing need for new protocols in mind, it seems better to have single modules for each incoming channel, although this leads to duplicated acceptance checks.
meillo@246 186
meillo@246 187
meillo@246 188 \subsubsection*{Outgoing channels}
meillo@246 189
meillo@246 190 Outgoing mail is commonly either sent using \SMTP, piped into local commands (for example \texttt{uucp}), or delivered locally by appending to a mailbox.
meillo@246 191
meillo@246 192 Outgoing channels are similar for \qmail, \postfix, and \name{sendmail X}: All of them have a module to send mail using \SMTP, and one for writing into a local mailbox. Local mail delivery is a job that requires root priveledge to be able to switch to any user in order to write to his mailbox. Modular \MTA{}s do not need \name{setuid root}, but the local delivery process (or its parent) needs to run as root.
meillo@246 193
meillo@246 194 As mail delivery to local users, is \emph{not} included in the basic job of an \MTA{}, why should it care about it? In order to keep the system simple and to have programs that do one job well, the local delivery job should be handed over to a specialist: the \name{mail delivery agent}. \NAME{MDA}s know about the various mailbox formats and are aware of the problems of concurrent write access and thelike. Hence handling the message and the responsiblity over to a \NAME{MDA}, like \name{procmail} or \name{maildrop}, seems to be the right way to go.
meillo@246 195
meillo@246 196 This means an outgoing connection that pipes mail into local commands is required. Other outgoing channels, one for each supportet protocol, may be designed like it was done in other \MTA{}s.
meillo@246 197
meillo@246 198
meillo@246 199
meillo@246 200 \subsubsection*{Mail queue}
meillo@246 201
meillo@246 202 Mail queues are probably used in all \mta{}s, excluding the simple forwarders. A mail queue is a essential requirement for \masqmail, as it is to be used for non-permanent online connections. This means, mail must be queued until a online connection is available to send the message.
meillo@246 203
meillo@246 204 The mail queue and the module to manage it are the central part of the whole system. This demands especially for robustness and reliability, as a failure here can lead to loosing mail. An \MTA\ takes over responsibility for mail in accepting it, hence loosing mail messages is absolutely to avoid. This covers any kind of crash situation too. The worst thing acceptable to happen is a mail to be sent twice.
meillo@246 205
meillo@246 206 \sendmail, \exim, \qmail, \name{sendmail X}, and \masqmail\ feature one single mail queue. \postfix\ has more of them.
meillo@246 207
meillo@246 208 \MTA\ setups that include content scanning tend to require two separate queues. To use \sendmail\ in such setups requires two independent instances, with two separate queues, running. \exim\ can handle it with special \name{router} and \name{transport} rules, but the data flow gets complicated. Hence an idea is to use two queues, \name{incoming} and \name{active} in \postfix's terminology, with the content scanning within the move from \name{incoming} to \name{active}.
meillo@246 209
meillo@246 210 \sendmail, \exim, \qmail, and \masqmail\ all use at least two files to store one message in the queue: one file contains the message body, another the envelope and header information. The one containing the mail body is not modified at all. \postfix\ takes a different approach in storing queued messages in an internal format within one file. \person{Finch} takes yet another different approach in suggesting to store the whole queue in one single file with pointers to separating positions \cite{finchFIXME}.
meillo@246 211 %fixme: check, cite, and think about
meillo@246 212
meillo@277 213 %fixme: discuss: filesystem vs. database
meillo@298 214 << \masqmail\ uses the filesytem to store the queue, storing the queue in a databases might improve the reliability through better persistence. >> %fixme
meillo@298 215
meillo@277 216 %fixme: what about the ``rule of repair''?
meillo@246 217
meillo@246 218
meillo@326 219 \subsection{Functional design}
meillo@246 220
meillo@246 221
meillo@246 222 \subsubsection*{Aliasing}
meillo@246 223
meillo@246 224 Where should aliases get expanded? They appear in different kind. Important are the ones available in the \path{aliases} file. Aliases can be:
meillo@285 225 \begin{enumerate}
meillo@246 226 \item a different local user (e.g.\ ``\texttt{bob: alice}'')
meillo@246 227 \item a remote user (e.g.\ ``\texttt{bob: john@example.com}'')
meillo@246 228 \item a list of users (e.g.\ ``\texttt{bob: alice, john@example.com}'')
meillo@246 229 \item a command (e.g.\ ``\texttt{bob: |foo}'')
meillo@285 230 \end{enumerate}
meillo@287 231 Addresses expanding to lists of users lead to more envelopes. Aliases changing the reciptients domain part may make the message unsuitable for a specific online route.
meillo@246 232
meillo@246 233 Aliasing is often handled in expanding the alias and reinjecting the mail into the system. Unfortunately, the mail is processed twice then; additionally does the system have to handle more mail this way. If it is wanted to check the new recipient address for acceptance and do all processing again, then reinjecting it is the best choice.
meillo@246 234
meillo@246 235
meillo@246 236
meillo@287 237 \subsubsection*{Route management}
meillo@246 238
meillo@287 239 %fixme: rework!!
meillo@246 240 One key feature of \masqmail\ is its ability to send mail out in different ways. The decision is based on the current online state and whether a route may be used for a message or not. The online state can be retrieved in tree ways, explained in \ref{sec:fixme}. A route to send is found by checking every available route for being able to transfer the current message, until one matches.
meillo@246 241
meillo@246 242 This functionality should be implemented in the module that is responsible to invoke one of the outgoing channel modules (for example the one for \SMTP\ or the pipe module).
meillo@246 243
meillo@246 244 \masqmail\ can rewrite the envelope's from address and the \texttt{From:} header, dependent on the outgoing route to use. This rewrite must be done \emph{after} it is clear which route a mail will take, of course, so this may be not the module where other header editing is done.
meillo@246 245 %fixme: see hafiz05 page 57: maybe put the rewriting into the sending module (like smx, exim, courier) (problem with archiving of all outgoing mail?)
meillo@246 246
meillo@246 247
meillo@246 248
meillo@246 249 \subsubsection*{Authentication}
meillo@246 250
meillo@246 251 One thing to avoid is being an \name{open relay}. Open relays allow to relay mail from everywhere to everywhere. This is a major source of spam. The solution is restricting relay\footnote{Relaying is passing mail, that is not from and not for the own system, through it.} access.
meillo@246 252
meillo@246 253 Several ways to restrict access are available. The most simple one is restrictiction by the \NAME{IP} address. No extra complexity is added this way, but static \NAME{IP} addresses are mandatory. This kind of restriction may be enabled using the operating system's \path{hosts.allow} and \path{hosts.deny} files. To allow only connections to port 25 from localhost or the local network \texttt{192.168.100.0/24} insert the line ``\texttt{25: ALL}'' into \path{hosts.deny} and ``\texttt{25: 127.0.0.1, 192.168.100.}'' into \path{hosts.allow}.
meillo@246 254
meillo@246 255 If static access restriction is not possible, for example if mail from locations with changing \NAME{IP} addresses wants to be accepted, some kind of authentication mechanism is required. Three common kinds exist:
meillo@246 256 \begin{enumerate}
meillo@246 257 \item \SMTP-after-\NAME{POP}: uses authenication on the \NAME{POP} protocol to permit incoming \SMTP\ connections for a limited time afterwards.
meillo@246 258 \item \SMTP authentication: is an extension to \SMTP. Authentication can be requested before mail is accepted.
meillo@246 259 \item Certificates: confirm the identity of someone.
meillo@246 260 \end{enumerate}
meillo@246 261
meillo@246 262
meillo@246 263
meillo@246 264 \subsubsection*{Encryption}
meillo@129 265
meillo@184 266 Electronic mail is very weak to sniffing attacks, because all data transfer is unencrypted. This concerns the message's content, as well as the email addresses in header and envelope, but also authentication dialogs that may transfer plain text passwords (\NAME{PLAIN} and \NAME{LOGIN} are examples). Adding encryption is therefor wanted.
meillo@184 267
meillo@184 268 The common way to encrypt \SMTP\ dialogs is using \name{Transport Layer Security} (short: \TLS, successor of \NAME{SSL}). \TLS\ encrypts the datagrams of the \name{transport layer}. This means it works below the application protocols and can be used by any of them\citeweb{wikipedia:tls}.
meillo@184 269
meillo@184 270 \TLS\ allows to create secure tunnels through which arbitrary programs can communicate. Hence one can add secure communication afterwards to programs without changing them. \name{OpenSSL} for example---a free implementation---allows traffic to be piped into a command; a secure tunnel is created and the traffic is forwarded through it. Or a secure tunnel can be set up between a local and a remote port; this tunnel can then be used by any application.
meillo@184 271
meillo@184 272 The \NAME{POP} protocol, for example, is good suited for such tunneling, but \SMTP\ is is not generally. Outgoing \SMTP\ client connections can be tunneled without problem---\masqmail\ already provides a configure option called \texttt{wrapper} to do so. Tunneling incomming connections to a server leads to problems with \SMTP. As data comes encrypted through the tunnel to the receiving host and gets then decrypted and forwarded on local to the port the application listens on. From the \MTA's view, this makes all connections appear to come from localhost, unfortunately. Figure \ref{fig:stunnel} depicts the data flow.
meillo@184 273
meillo@184 274 For incoming connections, \NAME{STARTTLS}---defined in \RFC2487---is what \mta{}s implement.
meillo@184 275
meillo@184 276 \masqmail\ is already able to encrypt outgoing connections, but encryption of incoming connections, using \NAME{STARTTLS} should be implemented. This only affects the \SMTP\ server module.
meillo@184 277
meillo@184 278
meillo@184 279
meillo@184 280
meillo@246 281
meillo@246 282 \subsubsection*{Spam prevention}
meillo@246 283
meillo@246 284 ---
meillo@287 285
meillo@287 286 Spam is a major threat nowadays and the goal is to reduce it to a bearable level (see section \ref{sec:swot-analysis}). Spam fighting is a war in which the good guys tend to lose. Putting too much effort there will result in few gain. Real success will only be possible with new---better---protocols and abandonning the weak legacy technologies. Hence \masqmail\ should be able to provide state-of-the-art spam protection, but not more.
meillo@287 287
meillo@246 288 ---
meillo@246 289
meillo@246 290 Spam is a major threat to email, as described in section \ref{sec:swot-analysis}. The two main problems are forgable sender addresses and that it is cheap to send hundreds of thousands of messages. Hence, spam senders can operate in disguise and have minimal cost.
meillo@246 291
meillo@287 292 As spam is not just a nuisance for end users but also for the infrastructure---the \mta{}s---by increasing the amount of mail messages. Thus \MTA{}s need to protect themself. Two different approaches are used:
meillo@246 293
meillo@287 294 \begin{enumerate}
meillo@287 295 \item Refusing spam during the \SMTP\ dialog. This is the way it was meant by the designers of the \SMTP\ protocol. They thought checking the sender and reciptient mail addresses would be enough, but as they are forgable it is not. More and more complex checks need to be done. Checking needs time, but \SMTP\ dialogs time out if it takes too long. Thus only limited time can be used, during the \SMTP\ dialog, for checking if a message seems to be spam. The advantage is that acceptance of bad messages can be simply refused---no responsibility for the message is taken and no further system load is added. See \RFC2505 (especially section 1.5) for detail.
meillo@246 296
meillo@287 297 \item
meillo@287 298 Checking for spam after the mail was accepted and queued. Here more processing time can be invested, so more detailed checks can be done. But, as responsibility for messages was taken by accepting them, it is no choice to simply delete spam mail. Checks for spam do not lead to sure results, they just indicate the possibility the message is unwanted mail. \person{Eisentraut} indicates actions to take after a message is recognized as probably spam \cite[pages 18--20]{eisentraut05}. The only acceptable one, for mail the \MTA\ is responsible for, is adding further or rewriting existent header lines. Thus all further work on the message is the same as for non-spam messages.
meillo@287 299 \end{enumerate}
meillo@246 300
meillo@301 301 Modern \MTA{}s use both techniques in combination. Checks during the \SMTP\ dialog tend to be implemented in the \mta\ to make it fast; checks after the message was queued are often done using external programs (\name{spamassassin} is a well known one). \person{Eisentraut} sees the checks during the \SMTP\ dialog to be essentiell: ``Ganz ohne Analyse w\"ahrend der \SMTP-Phase kommt sowieso kein \MTA\ aus, und es ist eine Frage der Einsch\"atzung, wie weit man diese Phase belasten m\"ochte.''\cite[page 25]{eisentraut05} (translated: ``No \MTA\ can go without analysis during the \SMTP\ phase anyway, but the amount of stress one likes to put on this phase is left to his discretion.'')
meillo@246 302
meillo@246 303 \NAME{DNS} blacklists (short: \NAME{DNSBL}) and \name{greylisting} are checks to be done before accepting the message. Invoking \name{spamassassin}, to add headers containing the estimated spam probability, is best to be invoked after the message is queued.
meillo@246 304
meillo@246 305
meillo@246 306
meillo@246 307
meillo@246 308 \subsubsection*{Virus checking}
meillo@246 309
meillo@246 310 Related to spam is malicous content (short: \name{malware}) like viruses, worms, trojan horses. They, in contrast to spam, do not affect the \MTA\ itself, as they are in the mail body. The same situation in the real world is post offices opening letters to check if they contain something that could harm the recipient. This is not a mail transport concern. Apart of not being the right program to do the job, the \MTA\---the one which is responsible for the recipient---is at a good position to do this work.
meillo@246 311
meillo@246 312 In any way should malware checking be done by external programs that may be invoked by the \mta. But using mail deliver and processing agents, like \name{procmail}, seem to be better suited locations to invoke content scanners.
meillo@246 313
meillo@246 314 A popular email filter framework is \name{amavis} which integrates various spam and virus scanners. The common setup includes a receiving \MTA\ which sends it to \name{amavis} using \SMTP, \name{amavis} processes the mail and sends it then to a second \MTA\ that does the outgoing transfer. \postfix\ and \exim\ can be configured so that one instance can work as both, the \MTA\ for incoming and outgoing transfer. A setup with \sendmail\ needs two separate instances running. It must be quarateed that all mail flows through the scanner.
meillo@246 315
meillo@246 316 A future \masqmail\ would do good to have a single point, where all traffic flows through, that is able to invoke external programs to do mail processing of any kind.
meillo@246 317
meillo@246 318
meillo@246 319 %AMaViS (amavisd-new): email filter framework to integrate spam and virus scanner
meillo@184 320 %\begin{verbatim}
meillo@246 321 %internet -->25 MTA -->10024 amavis -->10025 MTA --> reciptient
meillo@246 322 %| |
meillo@246 323 %+----------------------------+
meillo@246 324 %\end{verbatim}
meillo@184 325 %
meillo@246 326 %postfix and exim can habe both mta servises in the same instance, sendmail needs two instances running.
meillo@184 327 %
meillo@246 328 %MailScanner:
meillo@246 329 %incoming queue --> MailScanner --> outgoing queue
meillo@246 330 %
meillo@246 331 %postfix: with one instance possible, exim and sendmail need two instances running
meillo@184 332
meillo@184 333
meillo@246 334 %message body <-> envelope, header
meillo@246 335 %
meillo@246 336 %anti-virus: clamav
meillo@246 337 %postfix: via amavis
meillo@246 338 %exim: via content-scanning-feature called from acl
meillo@246 339 %sendmail: with milter
meillo@246 340 %procmail
meillo@246 341 %
meillo@246 342 %virus scanner work on file level
meillo@246 343 %amavis receives mail via smtp or pipe, splits it in its parts (MIME) and extracks archives, the come the virus scanners
meillo@246 344 %if the mail is okay, it goes via smtp to a second mta
meillo@184 345
meillo@246 346 %what amavis recognizes:
meillo@246 347 %- invalid headers
meillo@246 348 %- banned files
meillo@246 349 %- viruses
meillo@246 350 %- spam (using spam assassin)
meillo@246 351 %
meillo@246 352 %mimedefang: uses milter interface with sendmail
meillo@184 353
meillo@184 354
meillo@89 355
meillo@246 356 \subsubsection*{Archiving}
meillo@89 357
meillo@246 358 Mail archiving and auditability become more important as electronic mail becomes more important. Ability to archive verbatim copies of every mail coming into and every mail going out of the system, with relation between them, appears to be a goal to achieve.
meillo@194 359
meillo@246 360 \postfix\ for example has a \texttt{always\_bcc} feature, to send a copy of every mail to a definable reciptient. At least this funtionality should be given, although a more complete approach is preferable.
meillo@194 361
meillo@194 362
meillo@194 363
meillo@194 364
meillo@89 365
meillo@89 366
meillo@175 367
meillo@326 368 \subsection{Security design}
meillo@326 369
meillo@326 370 \subsubsection*{Sanitize mail}
meillo@326 371
meillo@326 372 Mail coming into the system often lacks important header lines. At least the required ones must be added from the \MTA. A good example is the \texttt{Message-Id:} header.
meillo@326 373
meillo@326 374 In \postfix, this is done by the \name{cleanup} module, which invokes \name{rewrite}. The position in the message flow is after coming from one of the several incoming channels and before the message is stored into the \name{incoming} queue. Modules that handle incoming channels may also add headers, for example the \texttt{From:} and \texttt{Date:} headers. \name{cleanup}, however, does a complete check to make the mail header complete and valid.
meillo@326 375
meillo@326 376 Apart from deciding where to sanitize the mail header, is the question where to generate the envelope. The envelope specifies the actual recipient of the mail, no matter what the \texttt{To:}, \texttt{Cc:}, and \texttt{Bcc:} headers tell. Multiple reciptients lead to multiple different envelopes, containing all the same mail message.
meillo@326 377
meillo@326 378
meillo@326 379
meillo@326 380
meillo@326 381
meillo@326 382
meillo@326 383
meillo@326 384
meillo@326 385
meillo@246 386
meillo@246 387 \subsection{The resulting architecture}
meillo@246 388
meillo@246 389 The result is a symetric design, featuring the following parts: Any number of handlers for incoming connections to receive mail and pass it to the module that stores it into the incoming queue. A central scanning module take mail from the incoming queue, processes it in various ways and puts it afterwards into the outgoing queue. Another module takes it out there and passes it to a matching transport module that transfers it to the destination. In other words, three main modules (queue-in, scanning, queue-out) are connected by the two queues (incoming, outgoing); on each end are more modules to receive and send mail---for each protocol one. Figure \ref{fig:masqmail-arch-new} depicts the new designed architecture.
meillo@246 390
meillo@246 391 \begin{figure}
meillo@246 392 \begin{center}
meillo@246 393 \includegraphics[width=\textwidth]{img/masqmail-arch-new.eps}
meillo@246 394 \end{center}
meillo@246 395 \caption{A new designed architecture for \masqmail}
meillo@246 396 \label{fig:masqmail-arch-new}
meillo@246 397 \end{figure}
meillo@246 398
meillo@246 399 This architecture is heavily influenced by the ones of \qmail\ and \postfix. Both have different incoming channels that merge in the module that puts mail into the queue; central is the queue (or more of them); and one module takes mail from the queue and passes it to one of the outgoing channels. Mail processing, in any way, is build in in a more explicit way than done in the other two. It is more similar to the \NAME{AR} module of \name{sendmail X}, which is the central point for spam checking.
meillo@246 400
meillo@246 401 Special regard was put on addable support for further mail transfer protocols. This appears to be most similar to \qmail, which was designed to handle multiple protocols.
meillo@246 402 %fixme: do i need all this ``quesses''??
meillo@246 403
meillo@246 404
meillo@246 405 \subsubsection*{Modules and queues}
meillo@246 406
meillo@246 407 The new architecture consists of several modules and two queues. They are defined in more detail now, and the jobs, identified above, are assigned to them. First the three main modules, then the queues, and afterwards the modules for incoming and outgoing transfer.
meillo@246 408
meillo@246 409
meillo@246 410 The \name{queue-in} module creates new spool files in the \name{incoming} queue for incoming messages. It is a process running in background, waiting for connections from one of the receiver modules. When one of them requests for a new spool file, the \name{queue-in} module opens one and returns a positive result. The receiver module then sends the envelope and message, which is written into the spool file by \name{queue-in}. If all went well, another positive result is returend.
meillo@246 411 %fixme: should be no daemon
meillo@246 412
meillo@246 413
meillo@246 414 The \name{scanning} module is the central part of the system. It takes spooled messages from the \name{incoming} queue, works on them, and writes them to the \name{outgoing} queue afterwards (the message is then removed from the \name{incoming} queue, of course). The main job is the processing done on the message. Headers are fixed and missing ones are added if necessary, aliasing is done, and external processing of any kind is triggered. The \name{scanning} module can run in background and look for new mail in regular intvals or signals may be sent to it by \name{queue-in}. Alternatively it can be called by \name{cron}, for example, to do single runs.
meillo@246 415
meillo@246 416
meillo@246 417 The \name{queue-out} module takes messages from the \name{outgoing} queue, queries information about the online connection, and then selects matching routes, creates envelopes for each recipient and passes the messages to the correct transport module. Successfully transfered messages are removed from the \name{outgoing} queue. This module includes some tasks specific to \masqmail.
meillo@287 418 %fixme: rework route selection
meillo@246 419
meillo@246 420
meillo@246 421 The \name{incoming} queue stores messages received via one of the incoming channels. The messages are in unprocessed form; only envelope data is prepended.
meillo@246 422
meillo@246 423
meillo@246 424 The \name{outgoing} queue contains processed messages. The header and envelope information is complete and in valid form.
meillo@246 425
meillo@246 426 \name{Receiver modules} are the communication interface between outside senders and the \name{queue-in} module. Each protocol needs a corresponding \name{receiver module} to be supported. Most popular are the \name{sendmail} module (which is a command to be called from the local host) and the \name{smtpd} module (which listens on port 25). Other modules to support other protocols may be added as needed.
meillo@318 427 %fixme: get invoked by inetd, or better ucspi-tcp (by bernstein) which can limit max number of concurrent connections. and includes tcp-wrappers functionality.
meillo@282 428
meillo@246 429
meillo@246 430 \name{Transport modules}, on the oppersite side of the system, are the modules to send outgoing mail; they are the interface between \name{queue-out} and remote hosts or local commands for further processing. The most popular ones are the \name{smtp} module (which acts as the \SMTP\ client) and the \name{pipe} module (to interface gateways to other systems or networks, like fax or uucp). A module for local delivery is not included, as it is in most other \MTA{}s; the reasons are described in FIXME.%fixme
meillo@246 431 Thus a \name{mail delivery agent} (like \name{procmail}) is to be used with the \name{pipe} module.
meillo@246 432
meillo@246 433
meillo@246 434
meillo@246 435 \subsubsection*{Inter-module communication}
meillo@246 436
meillo@246 437 Communication between modules is required to exchange data and status information. It is also called ``Inter-process communication'' (short: \NAME{IPC}), as modules are programs being part of a larger system, and processes are generally seen as programs in execution.
meillo@246 438
meillo@246 439 The connections between \name{queue-in} and \name{scanning}, aswell as between \name{scanning} and \name{queue-out} is provided by the queues, only sending signals to trigger instant runs may be useful. Communication between receiving and transport modules and the outside world are done using the specific protocol they do handle.
meillo@246 440
meillo@246 441 Left is only communication between the receiver modules and \name{queue-in}, and between \name{queue-out} and the transport modules. Data is exchanged done using \unix\ pipes and a simple protocol is used.
meillo@246 442
meillo@246 443 \begin{figure}
meillo@246 444 \begin{center}
meillo@273 445 \codeinput{input/ipc-protocol.txt}
meillo@273 446 %\includegraphics[scale=0.75]{img/ipc-protocol.eps}
meillo@246 447 \end{center}
meillo@246 448 \caption{State diagram of the protocol used for \NAME{IPC}}
meillo@246 449 \label{fig:ipc-protocol}
meillo@246 450 \end{figure}
meillo@246 451
meillo@246 452 % timing
meillo@246 453 One dialog consists of the four phases: connection attempt, acceptance reply, data transfer, success reply. The order is always the same. The connection attempt and data transfer are sent by the client process; replies are sent by the server process.
meillo@246 454 %fixme: split between header and data
meillo@246 455
meillo@246 456 % semantics
meillo@246 457 The connection attempt is simply opening the connection. This starts the dialog. A positive reply by the server leads to the data transfer, but a negative reply refuses the connection and resets both client and server to the state before the connection attempt. If the connection attempt was accepted, the client sends the data ending with a terminator sequence. When this terminator appears, the server process knows the complete data was transfered. The server process takes responsibility of the data in sending a positive success reply. A negative success reply resets both client and server to the state before the connection attempt.
meillo@246 458
meillo@246 459 The data transfered needs to be of specific format. Used is the same format in which messages are spooled in the mail queues. See the following section for details. %fixme: check if it is the following section
meillo@246 460 %fixme: split between header and data
meillo@246 461
meillo@246 462 % syntax
meillo@246 463 Data transfer is done sending plain text data. %fixme: utf8 ?
meillo@246 464 The terminator sequence used to indicate the end of the data transfer is a single dot on a line on its own. Line separators are the combination of \name{Carriage Return} and \name{Line Feed}, as it is used in various Internet protocols like \SMTP. Replys are one-digit numbers with \texttt{0} meaning success and any other number (\texttt{1}--\texttt{9}) indicate failure. %fixme: What are the octal values?
meillo@246 465 %fixme: split between header and data
meillo@246 466
meillo@246 467 Figure \ref{fig:ipc-protocol} is a state diagram for the protocol.
meillo@246 468
meillo@246 469
meillo@246 470
meillo@246 471 \subsubsection*{Spool file format}
meillo@246 472
meillo@246 473 The spool file format is basically the same as the one in current \masqmail: one file for the message body, the other for envelope and header information. The data file is stored in a separate data pool. It is written by \name{queue-in}, \name{scanning} can read it if necessary, \name{queue-out} reads it to generate the outgoing message, and deletes it after successful transfer. The header file (including the envelope) is written into the \name{incoming} queue. The \name{scanning} modules reads it, processes it, and writes a modified copy into the \name{outgoing} queue; the file in \name{incoming} is deleted then. \name{queue-out} finally takes the header file from \name{outgoing} to generate the resulting message. This data flow is shown in figure \ref{fig:queue-data-flow}.
meillo@246 474
meillo@246 475 \begin{figure}
meillo@246 476 \begin{center}
meillo@273 477 %\input{img/queue-data-flow.eps}
meillo@246 478 \end{center}
meillo@246 479 \caption{Data flow of messages in the queue}
meillo@246 480 \label{fig:queue-data-flow}
meillo@246 481 \end{figure}
meillo@246 482
meillo@246 483 The queue consists of three directories within the queue path. Two, named \name{incoming} and \name{outgoing}, for storing the header files; one, called \name{pool}, to store the message bodies. The files being part of one message share the same unique name. The header files internal structure can be the same as the one of current \masqmail.
meillo@246 484
meillo@246 485 Messages in queues are a header file in \name{incoming} or \name{outgoing} and a data file in \name{pool}. The header file owner's executable bit indicates if the file is ready for further processing: the module that writes the file into the queue sets the bit as last action. Modules that read from the queue can process messages with the bit set.
meillo@246 486
meillo@246 487 No spool files are modified after they are written to disk. Modifications to header files can be made by the \name{scanning} module in the ``move'' from \name{incoming} to \name{outgoing}---it is a create and remove, actually. Further rewriting can happen in \name{queue-out}, as well without altering the file.
meillo@246 488
meillo@246 489 Data files do not change at all within the system. They are written in default local plain text format. Required translation is done in the receiver and transport modules.
meillo@249 490 %fixme: why plain text and not db? -> simplicity
meillo@246 491
meillo@249 492 Mark spooled mail messages when processing of the writing module is finished: Either by setting the executable bit (like \postfix\ does), or by changing the owner (an approach for multiple masqmail users).
meillo@246 493
meillo@246 494
meillo@249 495 A sample header file. With comments in paranthesis.
meillo@246 496
meillo@261 497 \begin{quote}\footnotesize
meillo@246 498 \begin{verbatim}
meillo@246 499 1LGtYh-0ut-00 (backup copy of the file name)
meillo@246 500 MF:<meillo@dream> (envelope: sender)
meillo@246 501 RT: <user@example.org> (envelope: recipient)
meillo@246 502 PR:local (meta info: protocol)
meillo@246 503 ID:meillo (meta info: id/user/ip)
meillo@246 504 DS: 18 (meta info: size)
meillo@246 505 TR: 1230462707 (meta info: timestamp)
meillo@246 506 (following: headers)
meillo@246 507 HD:Received: from meillo by dream with local (masqmail 0.2.21) id
meillo@246 508 1LGtYh-0ut-00 for <user@example.org>; Sun, 28 Dec 2008 12:11:47 +0100
meillo@246 509 HD:To: user@example.org
meillo@246 510 HD:Subject: test mail
meillo@246 511 HD:From: <meillo@dream>
meillo@246 512 HD:Date: Sun, 28 Dec 2008 12:11:47 +0100
meillo@246 513 HD:Message-ID: <1LGtYh-0ut-00@dream>
meillo@246 514 \end{verbatim}
meillo@261 515 \end{quote}
meillo@246 516
meillo@246 517
meillo@246 518
meillo@246 519
meillo@246 520 \subsubsection*{Rights and permission}
meillo@246 521
meillo@246 522 The user set required for \qmail\ seems to be too complex. One special user, like \postfix\ uses, is more appropriate. \name{root} privilege and \name{setuid} permission is avoided as much as possible.
meillo@246 523
meillo@273 524 Table \ref{tab:new-masqmail-permissions} shows the suggested ownership and permissions of the modules.
meillo@246 525
meillo@246 526 \begin{table}
meillo@246 527 \begin{center}
meillo@271 528 \input{tbl/new-masqmail-permissions.tbl}
meillo@246 529 \end{center}
meillo@246 530 \caption{Ownership and permissions of the modules}
meillo@246 531 \label{tab:new-masqmail-permission}
meillo@246 532 \end{table}
meillo@246 533
meillo@273 534 These are the permissions and ownership used for the queue:
meillo@273 535 \codeinput{input/new-masqmail-queue.txt}
meillo@246 536
meillo@246 537
meillo@246 538
meillo@246 539
meillo@246 540
meillo@246 541 setuid/setgid or not?
meillo@246 542
meillo@246 543 what can crash if an attacker succeeds?
meillo@246 544
meillo@246 545 where to drop privelege?
meillo@246 546
meillo@246 547 how is which process invoked?
meillo@246 548
meillo@246 549 master process? needed, or wanted?
meillo@246 550
meillo@246 551 which are the daemon processes?
meillo@246 552
meillo@246 553
meillo@246 554
meillo@246 555
meillo@246 556
meillo@246 557
meillo@246 558
meillo@246 559 http://fanf.livejournal.com/50917.html %how not to design an mta - the sendmail command
meillo@246 560 http://fanf.livejournal.com/51349.html %how not to design an mta - partitioning for security
meillo@246 561 http://fanf.livejournal.com/61132.html %how not to design an mta - local delivery
meillo@246 562 http://fanf.livejournal.com/64941.html %how not to design an mta - spool file format
meillo@246 563 http://fanf.livejournal.com/65203.html %how not to design an mta - spool file logistics
meillo@246 564 http://fanf.livejournal.com/65911.html %how not to design an mta - more about log-structured MTA queues
meillo@246 565 http://fanf.livejournal.com/67297.html %how not to design an mta - more log-structured MTA queues
meillo@246 566 http://fanf.livejournal.com/70432.html %how not to design an mta - address verification
meillo@246 567 http://fanf.livejournal.com/72258.html %how not to design an mta - content scanning
meillo@246 568
meillo@246 569
meillo@246 570