Questioning the SMTP protocol

The SMTP protocol was a great invention and a key factor in the success of the Internet. Unfortunately, today’s unbearably high levels of spam show that it’s no longer entirely appropriate.

Today’s e-mail exchanges use SMTP(Simple Mail Transfer Protocol) to route messages from sender to receiver. This protocol has the advantage of being extremely simple to set up and use, and inexpensive, but this simplicity also makes it an ideal tool for spammers.

The STMP transaction depends essentially on the sender. He initiates the entire transaction, knows the sender’s identity in advance and sends the message asynchronously to the recipient. The decision is up to him, and he can decide to send as many e-mails as he likes to the recipient of his choice, whether or not the latter agrees to receive the message. Even if the final recipient decides to delete them without even looking at them, the fact remains that, thanks to the SMTP protocol, the messages have already been forwarded, and precious network resources have been consumed. There is no mechanism to temper this exclusive control by the shipper. In the early days of the Internet, when the network only connected trustworthy users, such a control system was not only superfluous, but also inadvisable, as it made the system cumbersome. Today, the nature of the Internet has changed, and what was a practical convenience has become a serious handicap.

Why do spammers keep flooding the world with their messages, when the vast majority of them go straight into the garbage can and are never even read? Simply the fact that it doesn’t cost them anything. A single positive response to a spam message is worth more than the cost of sending the whole batch. The SMTP protocol enables extremely significant economies of scale in e-mail transmission. Sending capacity is limited only by the computing power of the sending machine and the connectivity of the sender’s sending network. The sender just has to write and send the message, while the recipient spends more resources by having to be ready at all times to receive a message (since he can’t know in advance if and when a message will reach him), must then receive the message, process it, store it or delete it, even if he has no desire to receive the message. This can easily be explained by the desire to reduce loads as much as possible at a time when networks were less wide and slower. Today’s computers are more powerful and cheaper than ever, and spammers’ access to hundreds of thousands of zombie-like computers gives them access to a powerful, inexpensive and highly interconnected platform: Botnets.

Last but not least, it’s very easy for a spammer to disappear: in a very short space of time, he sends out a large number of spam messages and leaves the network; he hides behind a slew of zombie PCs; he changes locations and networks, either physically or via anonymizing proxies. Detection techniques based on IP addresses, such as Real Time Blacklists (RBLs), are therefore very limited in scope.

The changes or replacements to be made to SMTP must therefore be based on a protocol in which the recipient becomes the decision-maker, while the sender remains passive. Such protocols already exist: FTP or HTTP. It’s the Internet user who decides to download a file via FTP, not the other way around. In the case of SMTP, such a protocol would be identical to the way old pagers worked: the sender signals the recipient that he has a message to pass on. If the recipient is interested, he initiates the transaction by contacting the sender, who in turn sends the message. Such a system obviously has a number of drawbacks, not least because it complicates the life of the legitimate sender, who has to juggle between his numerous recipients and the storage of messages, or verify the identity of the recipient who contacts him. But the essential advantage of spam is that it can considerably reduce the volume of unsolicited messages passing through the Internet: even if spammers continue to operate, instead of the voluminous mass of messages we have today, we’ll have only very short, much lighter signalling messages. But there is a major problem: the very high false positive rate!