The Process of Security
By Bruce Schneier
I've been writing the CryptoRhythms column for this magazine for a little over a year now. When the editor and I sat down a couple months ago to talk about topics for 2000, I told him I wanted to expand the focus a bit from crypto-specific topics to broader information security subjects. So even though the column still falls under the CryptoRhythms banner, you can expect some (but not all) of this year's columns to address broader security issues that in some way incorporate cryptography. This month's article does just that, focusing on the process of security.
If we've learned anything from the past couple of years, it's that computer security flaws are inevitable. Systems break, vulnerabilities are reported in the press, and still many people put their faith in the next product, or the next upgrade, or the next patch. "This time it's secure." So far, it hasn't been.
Security is a process, not a product. Products provide some protection, but the only way to effectively do business in an insecure world is to put processes in place that recognize the inherent insecurity in the products. The trick is to reduce your risk of exposure regardless of the products or patches.
Will We Ever Learn?
Consider denial-of-service attacks. DoS attacks are some of the oldest and easiest attacks in the book. Even so, in February coordinated, distributed DoS attacks easily brought down several high-traffic Web sites, including Yahoo, eBay, Amazon.com and CNN.
Consider buffer overflow attacks. They were first talked about as early as the 1960s time-sharing systems suffered from the problem and were known by the security literati even earlier than that. In the 1970s, they were often used as a point of attack against early networked computers. In 1988, the Morris Worm exploited a buffer overflow in the Unix fingerd command: a very public use of this type of attack.
Today, over a decade after Morris and 35 years after these attacks were first discovered, you'd think the security community would have solved the problem of security vulnerabilities based on buffer overflows. Think again. Over two-thirds of all CERT advisories in 1998 were for vulnerabilities caused by buffer overflows. During an average week in 1999, buffer-overflow vulnerabilities were found in the RSAREF cryptographic toolkit (oops); HP's operating system; the Solaris operating system; and Microsoft IIS 4.0, Site Server 3.0, Windows NT and Internet Explorer. A recent study called buffer overflows the most common security problem.
Consider encryption algorithms. Proprietary secret algorithms are regularly published and broken. Again and again, the marketplace learns that proprietary secret algorithms are a bad idea. But companies and industries-such as Microsoft, the DVD consortium, cellular phone providers and so on-continue to choose proprietary algorithms over public, free alternatives.
Is Anyone Paying Attention?
Sadly, the answer to this question is, not really. Or at least, there are far fewer people paying attention than should be. And the enormous need for digital security products necessitates an enormous need for people to design, develop and implement them. This enormous need for people will be even greater than the number of skilled people, and so the percentage of people paying attention will get even smaller.
Most products that use security are not designed by anyone with security expertise. Even security-specific products are generally designed and implemented by people who have only limited security expertise. Security cannot be functionality-tested-no amount of beta testing will uncover security flaws--so the flaws end up in fielded products.
I'm constantly amazed by the kinds of things that break security products. I've seen a file encryption product with a user interface that accidentally saves the key in the clear. I've seen VPNs where the telephone configuration file accidentally allows a random person to authenticate himself to the server, or that allows one remote client to view the files of another remote client. There are a zillion ways to make a product insecure, and manufacturers manage to stumble on a lot of those ways again and again.
No one is paying attention because no one has to.
Computer security products, like software in general, have a very odd product quality model. It's completely unlike the quality-control process for an automobile or skyscraper...or even for a box of fried chicken. If you buy a product and get harmed because of a manufacturer's defect, you can sue...and you'll win. Car-makers can't get away with building cars that explode on impact; chicken shops can't get away with selling buckets of fried chicken with the odd rat mixed in. It just wouldn't do for building contractors to say things like, "Whoops. There goes another one. Sorry. But just wait for Skyscraper 1.1; it'll be 100 percent collapse-free!"
Software is different. It's sold without any claims whatsoever. Your accounts-receivable database can crash, taking your company down with it, and you have no claim against the software company. Your word processor can accidentally corrupt your files, and you have no recourse. Your firewall can turn out to be completely ineffectivehardly better than having nothing at all-and yet it's your fault. Microsoft fielded Hotmail with a bug that allowed anyone to read the accounts of 40 or so million subscribers, password or no password, and never bothered to apologize.
Software manufacturers don't have to produce a quality product because there is no liability if they don't. And the effect of this for security products is that manufacturers don't have to produce products that are actually secure, because no one can sue them if they make a bunch of false claims of security.
The upshot of this is that the marketplace doesn't reward real security. Real security is harder, slower and more expensive, both to design and to implement. Since the buying public has no way to differentiate real security from bad security, the way to win in this marketplace is to design software that is as insecure as you can possibly get away with.
Microsoft knows that reliable software is not cost-effective. According to studies, 90 to 95 percent of all bugs are harmless. They're never discovered by users, and they don't affect performance. It's much cheaper to release buggy software and fix the 5 to 10 percent of bugs people find and complain about.
Microsoft also knows that real security is not cost-effective. They get whacked with a new security vulnerability several times a week. They fix the ones they can, write misleading press releases about the ones they can't, and wait for the press fervor to die down (which it always does). And six months later, they issue the next software version with new features and all sorts of new insecurities, because users prefer cool features to security.
Fear Products; Embrace Process
There's no such thing as perfect security. Interestingly enough, that's not necessarily a problem. The credit card industry loses $10 billion to fraud per year in the U.S. alone, yet neither Visa nor MasterCard is showing any sign of going out of business. Shoplifting estimates in the U.S. are currently between $9.5 billion and $11 billion per year, but you never see "shrinkage" (as it's called) cited as the cause when a store goes out of business. Recently, I needed to notarize a document. That is about the stupidest security protocol I've seen in a long time. Still, it works fine for what it is.
Security does not have to be perfect, but the risks have to be manageable. The credit card industry understands this. They know how to estimate the losses due to fraud. Their problem is that losses from phone credit card transactions are about five times the losses from face-to-face transactions (when the card is presented). Losses from Internet transactions are about 10 times those of face-to-face transactions. Visa and MasterCard are pushing for Internet payment mechanisms precisely because the risks are getting worse.
My primary fear about cyberspace is that people don't understand the risks, and they're putting too much faith in technology's ability to obviate them. Products alone can't solve security problems.
The digital security industry is in desperate need of perceptual shift. Countermea- sures are sold as ways to avoid threats. Good encryption is sold as a way to prevent eavesdropping. A good firewall is marketed as a way to prevent network attacks. PKI is sold as trust management, so you can avoid mistakenly trusting people you really don't. And so on.
This type of thinking is completely backward. Security is old, older than computers. And the old-guard security industry thinks of countermeasures as ways to avoid risk. This distinction is enormous. Avoiding threats is black and white: either you avoid the threat, or you don't. Avoiding risk is continuous: there is some amount of risk you can accept, and some amount you can't.
Security processes are how you avoid risk. Just as businesses use the processes of double-entry bookkeeping, businesses need to use a series of security processes to protect their networks.
Security processes are not a replacement for products. Rather, they're a way of using security products effectively. They're a way to mitigate the risks. Network security products will have flaws; processes are necessary to catch attackers exploiting those flaws, and to fix the flaws once they become public. Insider attacks will occur; processes are necessary to detect the attacks, repair the damages and prosecute the attackers. Large systemwide flaws will compromise entire products and services (think digital cellphones, Microsoft Windows NT password protocols or DVD); processes are necessary to recover from the compromise and stay in business.
Here are two examples of how to focus on process in enterprise network security:
1. Watch for known vulnerabilities. Most successful network-security attacks target known vulnerabilities for which patches already exist. Why? Because network administrators either didn't install the patches, or because users reinstalled the vulnerable systems. It's easy to be smart about the former, but just as important to be vigilant about the latter. There are many ways to check for known vulnerabilities. Network vulnerability scanners such as Netect and SATAN test for them. Phone scanners like PhoneSweep check for rogue modems inside your corporation. Other scanners look for Web site vulnerabilities. Use these sorts of products regularly, and pay attention to the results.
2. Continuously monitor your network products. Almost everything on your network produces a continuous stream of audit information: firewalls, intrusion detection systems, routers, servers, printers, etc. Most of it is irrelevant, but some of it contains footprints from successful attacks. Watching it all is vital for security, because an attack that bypassed one product might be picked up by another. For example, an attacker might exploit a flaw in a firewall and bypass an IDS, but his attempts to get root access on an internal server will appear in that server's audit logs. If you have a process in place to watch those logs, you'll catch the intrusion in progress.
In these pages and elsewhere I have written pessimistically about the future of computer security. The future of computers is complexity, and complexity is anathema to security. The only reasonable thing to do is to reduce your risk as much as possible. We can't avoid threats, but we can reduce risk.
Nowhere else in society do we put so much faith in technology. No one has ever said, "This door lock is so effective that we don't need police protection, or breaking-and-entering laws." Products work to a certain extent, but you need processes in place to leverage their effectiveness.
Limit Privilege. Don't give any user more privileges than he absolutely needs to do his job. Just as you wouldn't give a random employee the keys to the CEO's office, don't give him a password to the CEO's files.
Secure the Weakest Link. Spend your security budget securing the biggest problems and the largest vulnerabilities. Too often, computer security measures are like planting an enormous stake in the ground and hoping the enemy runs right into it. Try to build a broad palisade.
Use Choke Points. By funneling users through choke points (think firewalls), you can more carefully secure those few points. Systems that bypass these choke points, like desktop modems, make security much harder.
Provide Defense in Depth. Don't rely on single solutions. Use multiple complementary security products, so that a failure in one does not mean total insecurity. This might mean a firewall, an intrusion detection system and strong authentication on important servers.
Fail Securely. Design your networks so that when products fail, they fail in a secure manner. When an ATM fails, it shuts down; it doesn't spew money out its slot.
Leverage Unpredictability. You know your network; your attacker doesn't. This is your big advantage. Make his job harder by disguising things, adding honey pots and booby traps, etc.
Enlist the Users. Security can't work if the users aren't on your side. Social engineering attacks are often the most damaging of any attack, and can only be defended against with user education.
Embrace Simplicity. Keep things as simple as absolutely possible. Security is a chain; the weakest link breaks it. Simplicity means fewer links.
DETECTION AND RESPONSE
Detect Attacks. Watch the security products. Look for signs of attack. Too often, valuable alerts from firewalls, servers and even IDSes are simply ignored.
Respond to Attackers. It's not enough to simply detect attacks. You need to close vulnerabilities when attackers find them, investigate incidents and prosecute attackers. We need to build a world where criminals are treated as such.
Be Vigilant. Security requires continuous monitoring; it's not enough to read a weekly report. Read about new attacks as soon as possible. Install all security patches and upgrades immediately.
Watch the Watchers. Audit your own processes. Regularly.
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.