The problem is self-signed certificates.

A CA is not a great solution:

Using a Certification Authority (CA) could solve the attack but at the same time introduces a new set of attack vectors:

1. The CA becomes a single point of failure. It becomes the juicy/high-value target for the attacker. Single point of failures are not good. Attractive targets are not good.

   Any person with access to the CA key can undetectably fake passports. Direct attacks, virus, misplacing the key by accident (the UK government is good at this!) or bribery are just a few ways of getting the CA key.

2. The single CA would need to be trusted by all governments. This is not practical as this means that passports would no longer be a national matter.

3. Multiple CA's would not work either. Any country could use its own CA to create a valid passport of any other country. Read this sentence again: Country A can create a passport data set of Country B and sign it with Country A's CA key. The terminal will validate and display the information as data from Country B.This option also multiplies the number of 'juicy' targets. It makes it also more likely for a CA key to leak.

   Revocation lists for certificates only work when a leak/loss is detected. In most cases it will not be detected.

So what's the solution? We know that humans are good at Border Control. In the end they protected us well for the last 120 years. We also know that humans are good at pattern matching and image recognition. Humans also do an excellent job 'assessing' the person and not just the passport. Take the human part away and passport security falls apart.

I’d love to have another blog even more specific - “Ok, that Assurance is Evidence Of What, Exactly?

Today he has a great article called:

What’s the matter with Risk Management?

And “in short, it’s everything.” It pretty much sums up why I had to grow to re-evaluate how our industry does risk, risk management, approaches controls & vulnerability and find a new way.   A couple of things jump out at me in reading Chris’ article:

1.)  Just because that Deming cycle sucks and is full of unknowns doesn’t mean “risk” doesn’t exist, nor that it isn’t of primary importance. Nor does it mean that in the absence of model & methodology, we won’t be “doing” risk analysis anyway - just in an ad hoc method and completely from “the gut”.

Our industry calls these unstructured risk analysis “Best Practices”, as it’s an easy and convenient way of sweeping the unknowns under the rug of bureaucracy and enforcing it via peer pressure.

2.)  What this “suckiness” does mean is that your model and methodology aren’t helping you. As Chris intimates, there is too much uncertainty in the inputs for his model (they are, in the language of Bayesians - too subjective to be useful priors).

Take for example how we might be approaching the “controls” part of our analysis.  Chris writes:

“2. What are the controls that we have to employ?
800-53, ISO 27001, PCI, etc.

Still kinda good, but we basically know that ISO is relatively voluntary and NIST supplies a control catalog and not policies. So here we have to take the control catalog, and mash our policies into it.”

I wouldn’t call this “kinda good” at all :)  These control catalogs only provide a hierarchy within which to look for evidence of  our ability to resist an attacker.  They are incapable of making any claim about the effectiveness of the controls when they are operated at 100% efficiency, or more importantly, what % efficiency our specific organization operates at.

Let’s use Chris Hayes’ Initech as our fictional example.

Initech has a control (a back door on a loading dock).  Now the locks on the door are 100% capable of locking the door.  This is different than saying that they are capable of frustrating all but the top 5% of lockpicking burgalars.  It is also diffferent than saying that in a sample of several “walk around audits” the doors are left open 20% of the time (they are not in compliance with policy 100% of the time).  Even worse, that 80% of the time the door is not propped open?  Yeah, tailgating is a known issue.

So we have several different variables here that we need to account for (and it’s just a door).  But the analogy stands that most “risk management” methodologies are “We have a door, yes/no?” And most GRC platforms, when asked for their “opinion” will simply say “door is needed” or, even worse, “a door policy is needed”.

3.)  Criticality and the Source of Value is all messed up in these Risk Management models.

Chris writes:

Someone wants me to tell them which boxes are more critical than others. This is mainly because of budgetary or operational reasons. To which I usually say “All of them, it is a system after all”.

This literally made me laugh out loud.  And this sort of “rate the firewall as Risk = 500 but rate the actual business application as Risk = 157″ thing is also endemic.  Now Chris is very smart here.  He correctly identifies that the value is tied to the business process the systems support, and not to a specific box.  Oh, we scan at the specific box level - but because of the nature of systemic failures - all the boxes in the process are inexorably interrelated.

One of the reasons I really like FAIR is that the losses are quantified (or qualified) based not on some amorphous value of the box or the process itself, but losses are linked to the actions that the threat will take. Take systems in a highly regulated industries as an example.  Usually the most probable losses aren’t due to system compromise per se, but in the disclosure the compromise causes (regulators are a threat source, after all).  But many “risk management” methodologies will say “online banking is worth $2 billion, the value of the systems is therefore $2 billion”.  And suddenly we’re telling executive management that there’s a 60% probability that they’ll lose $2 billion.

4.)  If the primary source of prior information for your “risk management” methodology is a vulnerability scanner - you’re doing it wrong.  Chris writes:

So we ran a scan and now we have a report. A snapshot in time to make all decisions. Where did these vulnerability ratings come from? Do I even know if my system is at risk? What if I spend my time on vulnerabilities that have no threat?

So first, my thoughts are that actual “vulnerability” must be a comparison of the force a threat can apply, and our ability to resist that force (this is a probability statement, btw).

Changing your thinking about vulnerability now helps us understand the problem in several new ways.  First, you can start to divorce yourself from the scanner.  After all, the scanner is simply providing you with current state information that is usually just relevant variance from policy. It doesn’t really tell you about real “weakness in a system” because the system is an interrelated mess of people, processes and IT assets.

5.)  Finally, most “risk management” approaches just *don’t* do a good job of helping us understand the how’s and why’s of managing risk. In the past, I’ve referred to these standards as really being “issue management” because they are at their heart, an act of discovery - a formal process around gathering prior information.  They are not, in and of themselves, capable of linking the issues discovered to the root cause.  And these root causes?  Yeah, they’re the things that create “risk”.  Not a threat, not a vulnerability, not the existence of an asset - the amount of risk that we have stems from our capability to manage it.

So Chris, I completely agree - but I wouldn’t give up yet.  There actually are a few of us who are focused on what you suggest:

Where to go from here: A fundamental revamp of how to deal with Risk. Where risk professionals focus on the treating the sickness and not the symptoms, and come up with some new success/actionable metrics.

Chris, there’s nothing I want to do more than that.

Blogger: Dan Blum

With the crisis in financial markets still unfolding, it is important to draw what lessons we can from the experience. Since the roots of the crisis lie in a monumental failure of risk management, it’s important to understand more about what happened, and then draw some parallels to our business risk management and  IT risk management situations.

The risk management failure in the housing market and on Wall Street had multiple interdependent dimensions:

• Mortgage lenders abandoned long standing prudent loan practices. They made too many loans that buyers might not be able to repay. Exotic instruments like ARMs, option ARMs, and interest only loans proliferated. In many cases, all pretense of lending standards were abandoned, so-called “liar loans” approved.
• Capital was grossly over-leveraged. Mortgage lenders and other financial services packaged loans into securities, which they sold to raise capital to support more lending. Real capital reserve requirements to back loans were reduced. Of course, if borrowers could not repay loans, all or parts of the derivative securities would become worthless.
• Risk was aggregated at Fannie Mae, Freddie Mac, and mortgage loan insurance companies. These companies bought or insured some mortgage loans, providing something of a backstop should loans fail. Government sponsored enterprises (GSEs) Fannie and Freddie in turn became over-leveraged and securities that they sold were in turn repackaged in the murky brew of mortgage-backed securities called collateralized debt obligations (CDOs) and other exotic instruments returning generous yields.
• Non-Caveat Emptor. Institutional wealth funds and financial services firms who should have known better bought securities that had been deliberately structured to obfuscate risk. They bought securities they didn’t understand with buried tranches of toxic subprime loans..

It was a great Ponzi scheme – one that kept working as long as housing prices were going up; the recipients of subprime loans could always flip that house to the next buyer. Everyone made money. As Chuck Prince of Citigroup famously put it during a July, 2007 interview: “So long as the music is playing, you’ve got to keep dancing. We’re still dancing.” But one month later, the music stopped. Since then, Citigroup and other financial institutions have taken massive writeoffs with more to come. Wall Street titans like Bear Sterns, Lehman Brothers, Merrill Lynch, and AIG have fallen or been bought out.

What can we learn from this risk management debacle?

As business risk managers and investors, we should ask questions like these:

• Does the executive incentive structure of the company encourage managers to dance around risk? Many Wall Street firms paid senior managers 5 times their salary in bonuses tied to annual growth alone.
• Is the company over-leveraged? Is it borrowing too much money and betting it on ventures with uncertain outcomes?
• Are financial models used for risk management realistic? Earlier, I described the mortgage market of the past few years as a Ponzi scheme, where risk management models must have assumed prices would keep rising. Unlike the dotcom boom whose demise many predicted, very few in the industry foresaw the sharp declines to come in housing prices and sales volumes. Historically, the U.S. housing market has been a steadily rising one, but on the other hand the 2000s saw unprecedented rates of price increases. In reality, what goes up must come down.
• Has your company’s risk council ever performed worst case scenario analysis and built adequate reserves? In the days before economics emerged as a would-be “hard” deterministic science, business leaders may have been more cautious, more aware of and more accepting of uncertainty. Events like the Great Tulip Bubble came once in decades or centuries – not every few years. Note that legendary investor George Soros has proposed a Theory of Reflexivity that, if true, helps explain the recent extremes of boom and bust cycles. This theory holds that market participants model market behaviors based on self-interest, and for a time, their manipulations change the reality of the market – until gravitational forces bring it back to earth. Has the music of ephemeral success played to the backbeat of deterministic-sounding economic models gone to your heads and infected your risk management models?
• Are cost cutting efforts pursued blindly? Outsourcing and other forays into treacherous global waters may be giving away the crown jewels. Smart companies cut costs, but they do it in smart ways. Smart companies think like intelligence agencies as they parcel out work to different partners with varying levels of dependability, and they check on those partners.

Risk management failures can also occur at the more technical level of IT security. As IT risk managers, we might ask questions like these:

• Are the accounting and financial systems your IT department supports under adequate control? As Fred Cohen wrote in one of our documents: “Many companies use computers to manage financial systems, and despite the Sarbanes-Oxley Act (SOX) claims about accounts being properly kept, there are many attacks on financial systems that remain. For example, most of the largest financial systems in the world running on common financial databases do not use double-entry bookkeeping and are thus susceptible to all manner of frauds by insiders.” We find it troubling that a prudent control dating back to the 12th century is going out of style in the name of convenience and cost cutting. Kind of like credit checking became anachronistic during the housing bubble, eh?
• Is the “separation” in your “separation of duty” (SoD) for real? Sure the SOX auditors are looking for SoD, and maybe you have different administrators with different accounts maintaining different systems or functions. But when they say Western civilization may be but one weak password from collapse they’re not lying. Look what happened to Sarah Palin’s email account! Weak and straggly SoD is a problem across all critical IT systems where deperimiterization and server consolidation may be bringing down protective barriers, identity management is weak, and strong process controls (e.g., where two people must sign on, one perform a critical operation such as backbone router reconfiguration, and the second observe) abandoned in the name of expediency.
• Are risks being aggregated to unacceptable levels in centralized control systems? There are many ways that risks aggregate within enterprise IT infrastructures as we pursue automation and cost cutting. Network risks aggregate when centralized domain name system control is implemented. Application risks aggregate when common infrastructure is shared among applications. And enterprises aggregate platform risks when they use low-assurance endpoints, authentication, and directory systems with single sign-on to access large numbers of resources and don’t separate high consequence systems.
• Non-caveat emptor: Has IT security really done the worst case consequence analysis, attack graphs, and vulnerability analysis to know when putting more eggs in a supposedly stronger basket aggregates risks to an unacceptable level? Or are you depending only on vendor claims about some black box appliance equivalent of a risk-obfuscated CDO security? Caveat emptor (buyer beware) again! (The good news is we’ll keep talking about promoting vendor and product rating systems so you don’t have to do all the detailed product analysis yourself, but that’s another post.)

There are many parallels between the monumental risk management failure in the financial markets, and the probable weaknesses in our day to day business risk management and IT risk management. Abandonment of prudent practices for profit; excessive leverage and centralization; ill-constructed risk analysis models; risk obfuscation; and a failure of caveat emptor seem to be common problems. Please take this as a wakeup call to sharpen up the risk management thinking, process, and execution.

Blogger: Trent Henry

Burton Group’s Catalyst Europe conference is just around the corner. With financial services industry failures at the top of everyone’s mind, now’s a great time to revisit how risk management shortcomings have tremendous impact on organizations of every kind. In a reprise of his insightful Catalyst North America talk, Nick Leeson will once again detail how inadequate controls (and foolish actions on his part) brought about the fall of Barings Bank. In addition, security conversations at Catalyst will include:

- How large enterprises are grappling with governance, risk, and compliance (and why “GRC” is actually a four-letter word)
- What large, distributed organizations are doing to create effective “security embassies”
- The role of metrics in managing protection and communicating with Management
- How information-centric security will unfold over the next five years

The session took place at the beginning of Partner Day. The “regular” conference sessions actually begin tomorrow. Today is spent focusing on partner issues and enablement.

The panel for this session included:

• Mark Chuang Group Manager, Product Marketing, VMware, Inc.
• Kenon Owens Staff Systems Engineer, VMware, Inc.

You have to remember that most of the Partners here are not vendors like ScienceLogic, but big and small shops that are selling IT, networking and now virtualization solutions into end-customer environments. For these guys, understanding what virtualization partner programs and tools are at NetApp, for example, is very useful. And many of these companies are already selling Microsoft software and surrounding services for Microsoft products. So if you’re VMware, what’s the message to these partners in the face of the Microsoft juggernaut?

Microsoft to partners: “You may not like to admit it, but you’re probably already in bed with us.”

VMware to partners: “Our hypervisor technology outperforms Hyper-V and Xen, especially at scale. And anyway, it’s not about the battle at the hypervisor. It’s about the V-services on top of the hypervisor – VMotion, Storage VMotion, DRS, etc.”

Interesting and what we all already know, or think we know. The scale issue is an interesting one – too soon for Hyper-V and who uses Xen? But also interestingly enough, no announcement or even talk about extending VMware management tools to other hypervisors. The point, as the VMware product marketing guy made a point of saying, is that the question they needed to answer used to be “Why Virtualization?” and now it’s “Why VMware?”.

One more tidbit – this survey run by VMware asking their customers:

What are the top 6 apps you are running on VMware today

• IIS
• Apache
• Active Directory
• SQL Server
• Sharepoint
• Exchange

That means, 5 of 6 are Microsoft applications. Certainly it makes it even more challenging for VMware to navigate a path here.

The change since 2004 – would have talked about why virtualize. And now why VMware. (Duh.)

Talking to partners – many of which already have a successful Microsoft business. How VMware enhances your existing Microsoft business.

Top 6 apps running on VMware today (5 of 6 are Microsoft applications)

• IIS
• Apache
• AD
• Sql server
• Sharepoint
• Exchange

Source: VMware survey

Esxi - VMware – true thin hypervisor; maximizes resources utilization (over 100% memory commitment – allows avg of 2:1 memory overcommit) – host system memory is usually the resource bottleneck – plus Advanced Scheduler runs VMs better under load and to a greater capacity (hard to show this part); performance acceleration – using binary translation (32bit), para-virtualization and Hardware Assist (for 64-bit)

(rvi – rapid virtualization indexing)

No parent partition that all hypervisors have to go through

Vs ms/xen

Parent partition – dom 0 => potentially problem at scale; i/o that could be a bottleneck

Hyper-v SPECjbb comparison

= 9 vms on VMware and hyper-v hypervisors

Outperform (CPU) by 50% - general purpose scheduler isn’t able to keep up? “got to be”

(cpu only test)

Also used VMmark – to demonstrate again that VMware is performance tuned and designed to run at scale vs Hyper-V

Size Does Matter:

Vmware ESXi: 32MB

Hyper-v – 2.6 GB

Xen – 1.2 GB

Hyper-V uses Microsoft Server Core – so the last two Patch Tuesdays had to make changes to Server Core (nothing to do with Hyper-V) but service interruption for Hyper-V.

VMware VMsafe – “Provides an unprecedented level of security” “virtual is more secure than Real” (uh oh – clearly didn’t read about the

*****************

VMware TEST:512 mb vms on server w/ 4gb ram –

7 vms - xensource (w/no memory overcommit)

6vms – hyper-v before error (w/no memory overcommit)

14vms - w/memory overcommit and management

Running sql io sim – heavy workloads

TCO – not just license; now ESXi is free – so hardware

809 - ESXi

871 – vi3 foundation ($995)

1168- vi3 enterprise ($5750)

1621 – hyper-v – 2x cost because of hw

Xen – 1618

Memory overcommit (89% in production vs. test/dev)

Survey – 37% of respondents at 2:1 RATIO OR HIGHER; real average is around 1.8: 1

*********************

This guy Mark sounds like a used car salesman:

“Always On, On Demand Data Center”

Hypervisor is very important but what is more important are the v-services on top of this. Manage shared, pooled resources. “Value Above the Hypervisor”

How does all this save “your customers” $$?

VMotion – saves cost on planned maintenance: no more overtime, no more time scheduling maintenance windows (see cost framework below)

10 (# of servers) x 6 (@ of updates) x [ (overtime cost 2hrs x $150/hr) + (scheduling downtime # of apps per server 15 x time spend scheduling per app 0.75 hr x $50/hr)] = $58,500

Same thing with using VMware Storage VMotion

Overtime cost + scheduling downtime + planning move + alternative tool cost - $68,750 (2.5 TeraBytes)

The Value of High Availability

- cost of lost business, lost work

- cost of lost productive time

4 hours of downtime x # of users per vm 10 x number of vms per host 15 x cost of user productive time $50/hr x failures per year in 10-host cluster 2 = $60K

(10 servers, 150 vms)

SAVINGS (using enterprise version)

Update management 149,760

HA 60K

DRS, VMotion Storage VMotion 187,250

808,259 – hw, power cooling, etc.

Freud wrote that dreams are the "Royal Road" to the unconscious. Messaging Systems and mainframes can be the Royal Road for a malicious user to attack a banking system, unless the security thinking moves from a "claim by claim" thinking to thinking about the full transaction.

In the case of phishing, it is relatively clear. The developers believe the PKI book. The PKI people believe in the efficacy of digital signatures to prove stuff. The cryptographers believe in the perfection of mathematics, and the security world believes in the completeness of their own learning. They are all wrong, but only at the large level of generalisations, not at the detailed level of particular claims. Any one of the claims, in isolation can be shown to be true. But, generalising these brittle claims to be solid building blocks is a completely different question. Few of the claims are strong enough to partake in a general model without severe support; the general model of secure browsing is the best evidence of how it is secure in name only.

How then is it built? By accident or by design, a series of claims meet together in a holy ring of righteous architecture. Each of the proponents claim loudly that their part is strong, but the ring has no strength. Eventually, one of the claims in the links is broken. For phishing, the browsers never did have the potential to show authenticity; not only did they not have the security strength to do it (c.f., Skype v. CSRF), they didn't even do it in practice (recall the lost padlock?), and their recent efforts to show authenticity (c.f. colour debate) reveal how far they are from understanding even the goal, let alone the implementation. Once that link was broken, and money was made, all the others revealed their weaknesses, as crooks systematically worked to breach the lot.

If we look at the wider financial collapse, now underscored by the nationalisation of the worlds biggest financiers of mortgages ($ 5.3 trillion.... or is it $ 5.4 ?), we see the same pattern. The bankers believed in their product. The originators believed in their origination, the securitizers believed in their free market and accurate price, and the holders believed in the assets. The CDO, the subprime, the other 100 special names, each was a contract. Each was clear in and of itself. But, when placed end-to-end, in a line, with a bunch of other agreements, the claims that were good in isolation were not strong enough to participate in the super-claim made of the overall edifice.
The financial system was built like a bridge; each piece rested on the previous one. And then, the clever architects bent the bridge around ... and around again, until the first piece met the last. The elegant keystone of finance was to finally lift up the first one to rest on the last.

Thus, the banks themselves invested their capital in their own product.

So, Anton Security Tip of the Day #16: Virtually Screwed - Journey Into VMWare ESX Log Analysis

CISecurty guide for VMWare (here) and DISA STIG for virtual machines (here) both mandate collection and analysis of VM platform logs; none goes into enough details on what to look for in logs. Let's try to shed some light on security-focused log analysis of VMWare ESX v. 3.x logs.

First, at least until ESXi becomes the default choice, one needs to keep in mind that ESX as "Linux-inside" and thus diving into /var/log will not reveal any "alien technology" (well, not much :-)). However, one of the most useful logs is /var/log/hostd.N which is not a descendant of Linux standard logs. Extensive VM event records are written into this file.

Let's focus on various types of logins to the ESX platform and identify logs that indicate a successful and failed attempts to log in. Here are a few useful examples to analyze:

Successful logins:

• May 30 09:20:42 esx2 su(pam_unix)[9405]: session opened for user root by jhonny(uid=1626)

This is a classic Linux root login message; you can watch for these by searching VMWare ESX logs for "session AND opened AND user AND root."  Notice the user name of the user who switched to root.

• May 30 09:20:34 esx2 sshd(pam_unix)[9364]: session opened for user jhonny by (uid=0)

This is also a classic Linux message for a normal (non-root) user login.

• [2008-05-25 06:57:48.774 'ha-eventmgr' 111639472 info] Event 40645 : User jhonny@1.1.1.1 logged in

This is a VMWare -specific application login to ESX. You can track such events by username, by event ID or by keywords "event AND logged AND user" (if you are using search)

Failed logins:

• May 30 09:20:31 esx2 sshd[9356]: Failed password for jhonny from 1.1.1.1 port 54773 ssh2

Another classic Linux message from the ESX system; a failure to login due to incorrect password.

• May 27 12:06:59 esx2 sshd[4756]: Failed password for illegal user jonny from 1.1.1.1 port 30594 ssh2

A message indicating a failure to login due to incorrect username (note a typo).

• May 25 07:03:48 esx1 sudo:     jhonny : 3 incorrect password attempts ; TTY=pts/0 ; PWD=/var/log ; USER=root ; COMMAND=/bin/bash

This ESX Linux platform message should also be familiar to Linux/Unix admins: it indicates multiple sudo password failures; look for such messages in the logs.

BTW, do you need to be reminded to track NOT only failed, but also successful login events?!

Overall, you must prepare for the future by learning to analyze  VMWare logs, just like you handled "legacy OS", such as Linux/Unix and Windows.

As I said before, I am tagging all the tips on my del.icio.us feed; here is the link: All Security Tips of the Day.

It's basically what you would expect: Appoint a national cyber security advisor, invest in math and science education, establish standards for critical infrastructure, spend money on enforcement, establish national standards for securing personal data and data-breach disclosure, and work with industry and academia to develop a bunch of needed technologies.

I could comment on the plan, but with security the devil is always in the details -- and, of course, at this point there are few details. But since he brought up the topic -- McCain supposedly is "working on the issues" as well -- I have three pieces of policy advice for the next president, whoever he is. They're too detailed for campaign speeches or even position papers, but they're essential for improving information security in our society. Actually, they apply to national security in general. And they're things only government can do.

One, use your immense buying power to improve the security of commercial products and services. One property of technological products is that most of the cost is in the development of the product rather than the production. Think software: The first copy costs millions, but the second copy is free.

You have to secure your own government networks, military and civilian. You have to buy computers for all your government employees. Consolidate those contracts, and start putting explicit security requirements into the RFPs. You have the buying power to get your vendors to make serious security improvements in the products and services they sell to the government, and then we all benefit because they'll include those improvements in the same products and services they sell to the rest of us. We're all safer if information technology is more secure, even though the bad guys can use it, too.

Two, legislate results and not methodologies. There are a lot of areas in security where you need to pass laws, where the security externalities are such that the market fails to provide adequate security. For example, software companies who sell insecure products are exploiting an externality just as much as chemical plants that dump waste into the river. But a bad law is worse than no law. A law requiring companies to secure personal data is good; a law specifying what technologies they should use to do so is not. Mandating software liabilities for software failures is good, detailing how is not. Legislate for the results you want and implement the appropriate penalties; let the market figure out how -- that's what markets are good at.

Three, broadly invest in research. Basic research is risky; it doesn't always pay off. That's why companies have stopped funding it. Bell Labs is gone because nobody could afford it after the AT&T breakup, but the root cause was a desire for higher efficiency and short-term profitability -- not unreasonable in an unregulated business. Government research can be used to balance that by funding long-term research.

Spread those research dollars wide. Lately, most research money has been redirected through DARPA to near-term military-related projects; that's not good. Keep the earmark-happy Congress from dictating how the money is spent. Let the NSF, NIH and other funding agencies decide how to spend the money and don't try to micromanage. Give the national laboratories lots of freedom, too. Yes, some research will sound silly to a layman. But you can't predict what will be useful for what, and if funding is really peer-reviewed, the average results will be much better. Compared to corporate tax breaks and other subsidies, this is chump change.

If our research capability is to remain vibrant, we need more science and math students with decent elementary and high school preparation. The declining interest is partly from the perception that scientists don't get rich like lawyers and dentists and stockbrokers, but also because science isn't valued in a country full of creationists. One way the president can help is by trusting scientific advisers and not overruling them for political reasons.

Oh, and get rid of those post-9/11 restrictions on student visas that are causing (.pdf) so many top students to do their graduate work in Canada, Europe and Asia instead of in the United States. Those restrictions will hurt us immensely in the long run.

Those are the three big ones; the rest is in the details. And it's the details that matter. There are lots of serious issues that you're going to have to tackle: data privacy, data sharing, data mining, government eavesdropping, government databases, use of Social Security numbers as identifiers, and so on. It's not enough to get the broad policy goals right. You can have good intentions and enact a good law, and have the whole thing completely gutted by two sentences sneaked in during rulemaking by some lobbyist.

Security is both subtle and complex, and -- unfortunately -- it doesn't readily lend itself to normal legislative processes. You're used to finding consensus, but security by consensus rarely works. On the internet, security standards are much worse when they're developed by a consensus body, and much better when someone just does them. This doesn't always work -- a lot of crap security has come from companies that have "just done it" -- but nothing but mediocre standards come from consensus bodies. The point is that you won't get good security without pissing someone off: The information broker industry, the voting machine industry, the telcos. The normal legislative process makes it hard to get security right, which is why I don't have much optimism about what you can get done.

And if you're going to appoint a cyber security czar, you have to give him actual budgetary authority -- otherwise he won't be able to get anything done, either.

This essay originally appeared on Wired.com.

It's basically what you would expect: Appoint a national cyber security advisor, invest in math and science education, establish standards for critical infrastructure, spend money on enforcement, establish national standards for securing personal data and data-breach disclosure, and work with industry and academia to develop a bunch of needed technologies.

I could comment on the plan, but with security the devil is always in the details -- and, of course, at this point there are few details. But since he brought up the topic -- McCain supposedly is "working on the issues" as well -- I have three pieces of policy advice for the next president, whoever he is. They're too detailed for campaign speeches or even position papers, but they're essential for improving information security in our society. Actually, they apply to national security in general. And they're things only government can do.

One, use your immense buying power to improve the security of commercial products and services. One property of technological products is that most of the cost is in the development of the product rather than the production. Think software: The first copy costs millions, but the second copy is free.

You have to secure your own government networks, military and civilian. You have to buy computers for all your government employees. Consolidate those contracts, and start putting explicit security requirements into the RFPs. You have the buying power to get your vendors to make serious security improvements in the products and services they sell to the government, and then we all benefit because they'll include those improvements in the same products and services they sell to the rest of us. We're all safer if information technology is more secure, even though the bad guys can use it, too.

Two, legislate results and not methodologies. There are a lot of areas in security where you need to pass laws, where the security externalities are such that the market fails to provide adequate security. For example, software companies who sell insecure products are exploiting an externality just as much as chemical plants that dump waste into the river. But a bad law is worse than no law. A law requiring companies to secure personal data is good; a law specifying what technologies they should use to do so is not. Mandating software liabilities for software failures is good, detailing how is not. Legislate for the results you want and implement the appropriate penalties; let the market figure out how -- that's what markets are good at.

Three, broadly invest in research. Basic research is risky; it doesn't always pay off. That's why companies have stopped funding it. Bell Labs is gone because nobody could afford it after the AT&T breakup, but the root cause was a desire for higher efficiency and short-term profitability -- not unreasonable in an unregulated business. Government research can be used to balance that by funding long-term research.

Spread those research dollars wide. Lately, most research money has been redirected through DARPA to near-term military-related projects; that's not good. Keep the earmark-happy Congress from dictating (.pdf) how the money is spent. Let the NSF, NIH and other funding agencies decide how to spend the money and don't try to micromanage. Give the national laboratories lots of freedom, too. Yes, some research will sound silly to a layman. But you can't predict what will be useful for what, and if funding is really peer-reviewed, the average results will be much better. Compared to corporate tax breaks and other subsidies, this is chump change.

If our research capability is to remain vibrant, we need more science and math students with decent elementary and high school preparation. The declining interest is partly from the perception that scientists don't get rich like lawyers and dentists and stockbrokers, but also because science isn't valued in a country full of creationists. One way the president can help is by trusting scientific advisers and not overruling them for political reasons.

Oh, and get rid of those post-9/11 restrictions on student visas that are causing (.pdf) so many top students to do their graduate work in Canada, Europe and Asia instead of in the United States. Those restrictions will hurt us (.pdf) immensely in the long run.

Those are the three big ones; the rest is in the details. And it's the details that matter. There are lots of serious issues that you're going to have to tackle: data privacy, data sharing, data mining, government eavesdropping, government databases, use of Social Security numbers as identifiers, and so on. It's not enough to get the broad policy goals right. You can have good intentions and enact a good law, and have the whole thing completely gutted by two sentences sneaked in during rulemaking by some lobbyist.

Security is both subtle and complex, and -- unfortunately -- it doesn't readily lend itself to normal legislative processes. You're used to finding consensus, but security by consensus rarely works. On the internet, security standards are much worse when they're developed by a consensus body, and much better when someone just does them. This doesn't always work -- a lot of crap security has come from companies that have "just done it" -- but nothing but mediocre standards come from consensus bodies. The point is that you won't get good security without pissing someone off: The information broker industry, the voting machine industry, the telcos. The normal legislative process makes it hard to get security right, which is why I don't have much optimism about what you can get done.

And if you're going to appoint a cyber security czar, you have to give him actual budgetary authority -- otherwise he won't be able to get anything done, either.

---

Bruce Schneier is chief security technology officer of BT, and author of Beyond Fear: Thinking Sensibly About Security in an Uncertain World.