![]() |
|
|||||||||||||||
|
||||||||||||||||
|
| Keeping e-mail secure | ||||
| Some useful tips for keeping snoopers out and sensitive data in
The hot debate surrounding the discovery of the U.S. FBI's Carnivore system has one beneficial effect: Companies have been given a huge heads-up that e-mail can be snooped without warning, notice, or even detection. What the FBI can do, so can a "black hat" performing industrial espionage. This article addresses common-sense approaches for preventing your electronic mail from compromising your company or your business. In particular, it discusses the new addition to the mail protocols that enhances security of electronic mail. Background There is no rocket science in the construction of Carnivore. The core of the thing is a commercially available network packet monitor/analyzer -- or "sniffer" -- a network diagnostic tool that's used by network administrators everywhere. During Congressional hearings about Carnivore, network administrators testifying to Congress demonstrated just how easy it is to tap into electronic mail using these tools without disturbing servers or networks -- without setting off alarms of any kind. A number of vendors sell powerful sniffers, but anyone can use Open BSD or Linux for free, along with readily available programs and libraries, to do the same job just for the price of an entry-level personal computer and the effort of downloading the source and compiling the packages. The U.S. FBI is just a new member of a large group abusing diagnostic tools for non-diagnostic purposes (such as information espionage). An espionage agent using a packet sniffer to snoop Internet activity does so with a specific purpose in mind. Like any data mining activity, the "take" has to be managed with that purpose in mind, lest the poor agent succumb to being buried in the stuff that has been snooped or starved of the very information they were looking for. The snooping methods can be broken down into three classes. The most voluminous method is full-content snooping, where the full text of messages is intercepted and stored. The next level up in abstraction is detailed traffic monitoring, where the sending and receiving addresses are captured, and perhaps some characteristics of the message (length, message type, or other easily detected attributes) are stored. The most basic level of monitoring is site-to-site (or site-level) traffic monitoring, where the traffic volume between IP addresses is monitored. Full content snooping Full-text capture is also the easiest snooping to block: Just encrypt the contents of electronic mail with any number of packages. Many people have been using Phil Zimmerman's PGP (PrettyGoodPrivacy), currently available for free download at Network Associates, Inc.'s Web site (see Resources). Commercial versions are also available at NAI. Other vendors have made encryption packages available. The S/MIME standards are described in RFCs 2631 through 2634 (version 3) and an earlier effort in RFCs 2311 and 2312. (See Resources.) Because there are several incompatible encryption products available, you should plan your encryption implementation carefully. For example, you should be sure that all parties share a common encryption algorithm so that a message encrypted on one system can be decrypted on the others. Get it to work before you need it. The next question is how strong an encryption you should use. The goal of encryption is not to absolutely protect data; rather, it is to increase the time and cost of "cracking the code" to the point that the return on the effort is too small to be worth expending the effort. For public-key cryptography, a 512-bit key is considered less effective, while a 2,048-bit key is considered to have a significant life. Detailed traffic monitoring Law enforcement puts its faith in pen-register orders and trap-and-trace orders. The legal barrier to getting a court order for a full ("Title III") wiretap is a high one in the courts: The law enforcement officers have access to everything said on the telephone line, so the requesting authority needs to show an overriding need to invade. The barrier for pen-register and trap-and-trace orders is much lower because the information gleaned from such orders is not considered by the courts to be "content," but rather "addressing information." The courts have already extended this concept to Internet transmissions. (Whether the implementation of Carnivore captures content beyond that allowed by Court order -- as a number of privacy advocacy groups claim -- is the core of the current debate, and is beyond the scope of this article.) Encryption of the body of the message can't protect the addressing information: The protocols used for transferring mail separate the address from the body, much as the outside of a mail envelope exposes the address information while the content is tucked inside. There's even a question whether capturing address information is considered "wiretapping." In the United States, courts have made a distinction between content capture and address ("digit capture" in the case of the telephone) in terms of the type of search warrant required. Because of this bifurcation, you (or your company) may find yourself hampered in prosecuting an espionage agent who clearly just "reads addresses" and not content. Content can be important to a industrial espionage agent only insofar as it classifies whether the message is plain text or cipher text. If a company is in the habit of encrypting "delicate" messages, the agent can measure separately the number, sender, and recipient of such messages and derive useful information from that. One way to reduce the effectiveness of such classification is to encourage widespread use of encryption, even for less sensitive material. This robs the espionage agent of even the most crude message sorting method. Site-level traffic monitoring Defeating this sort of monitoring is easy. For leased-line circuits, consider using an external compression and encryption box on each end of the circuit. For T1 circuits, not only does this protect your traffic, but it can increase the data rate from 1.544 Mbps to over 3 Mbps. Encryption, encryption, encryption Virtual private networking (VPN) is the process of setting up a "tunnel" through the Internet from one point to another. The network is private because encryption is used at the two ends, to hide the information passing through the public medium (i.e., the Internet) from prying eyes. Cisco and Lucent build VPN capability into their router products, while numerous other manufacturers offer appliances. Microsoft includes VPN capability in its Windows product line. Mail sent through a VPN tunnel is effectively protected from snoopers. Other applications that use the tunnel -- such as Web browsing, file transfers, and IRC chats -- are protected as well. Indeed, when a VPN is used as a general-purpose virtual link, the VPN is transparent to all TCP/IP applications. The tunnel concept has been extended to mail transfers generally to increase the security of all mail traffic. RFC 2487 describes a method of extending the Simple Mail Transfer Protocol (SMTP) to use Transport Layer Security (TLS -- also known as Secure Socket Layer, or SSL) tunnels. This means that mail sent from a TLS-aware mail client to a mail server, or between mail servers, is encrypted by default -- addresses and content. The mail client doesn't have to understand how to decrypt individual mail messages in such a system. Further, the addressing information is hidden from eavesdroppers. How does it work? The handshake between mail transfer agents, the software that actually exchanges mail from server to server or from the client to the first server, includes a special indicator that says, in essence, "I know how to do encrypted transfer." The two pieces of software exchange this special token, build an encrypted channel, and then proceed as they always do. The key is that all information -- address and content -- is transmitted in an encrypted channel. The encryption is as strong as the two servers allow, which means that all mail ends up being encrypted between two TLS-capable mail transfer agents, and not just those messages considered delicate. The addresses are hidden as well, so that if the encryption isn't cracked in the virtual channel, the only information available to the espionage agent is site-to-site traffic volume, with no indication of what percentage of the traffic is sensitive. The certificate exchange system used in TLS-enhanced SMTP can also protect against forged headers that cause people to leak secrets. Users can check to see if the source of the e-mail follows a path that is legitimate for that person. The servers known to provide RFC 2487 capability are Sendmail version 8.11 and Microsoft Exchange Server 5.5. Qmail by Dan Bernstein augmented with a patch by Frederik Vermeulen is an experimental implementation. TLS-enhanced IMAP, POP3, and ACAP Today, there are no commercial mail clients for Windows that implement RFC 2595. Expect to see new products by the end of the year from Qualcomm/Eudora and from Rit Labs, just to name two. Security is more than just eavesdropping Encryption techniques are only one way of protecting your data. You also have to pay attention to physical security, access control, and document storage to ensure that no espionage agent can gain access to your data. Conclusion The future? TLS-enhanced mail can only grow now that the bellweather mail transfer agent, Sendmail, has the capability built-in. Other mail transfer agent programs will need to be enhanced in the same way. That further reduces the possibility of casual eavesdropping as time goes on. If you are breaking the law, don't think for a moment that these techniques will hide your actions. Law enforcement has other ways of catching and convicting you.
| |||||||||||||||
| About IBM | Privacy | Terms of use | Contact |