Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Network Intrusion Detection, Third Edition.pdf
Скачиваний:
222
Добавлен:
15.03.2015
Размер:
2.58 Mб
Скачать

threat. Vulnerabilities are the gateways by which threats manifest themselves.

Defining the Threat

"Umm, I wouldn't go there if I were you". "Why not?"

"Bad things will happen to you if you go there." "What bad things?"

"Bad things."

This is not a compelling scenario, true? Most of us would not be persuaded by it. Imagine giving a similar pitch to management: If you don't fund an intrusion-detection system, bad things will happen to us.

We need to define and quantify bad things:

What things?

How bad?

How likely they are to occur or repeat?

How do you know?

What support do you have for your answer?

So for each threat we can define and enumerate, we need to answer these questions.

How Bad—Impact of Threat

In the end, risk is evaluated in terms of money. This is true even if life is lost; in the case of loss of life, it might be a lot of money. For any threat we have defined, we take the value of the assets at risk and multiply that by how exposed they are. This yields the expected loss if we were to get clobbered by the threat. This is called the single loss expectancy (SLE) and the

formula to calculate SLE is as follows:

Asset value x exposure factor = SLE

The exposure factor is an estimate, ranging from 0 percent to 100 percent of our loss of the asset. Consider the following calculation, the threat of a nuclear bomb exploding just above a

small town whose total assets are worth 90 million dollars:

Example Nuclear bomb/small town ($90M x 100% = $90M)

Now let's bring it home. I have already mentioned that when I have conducted vulnerability scans of sites with UNIX computers I have found a number of systems with the tooltalk vulnerability. Can we apply this formula to these? First, we have to define the threat. Suppose we are a Class C site. The threat is a malicious attacker who gains root, exploits any trust models, encrypts the file systems, and holds the computers ransom for $250,000. The attacker scans the net and finds six vulnerable systems. The buffer-overflow attack quickly yields root. After exploiting the trust models of these systems, our attacker is able to root compromise four additional systems and therefore encrypt the disks of 10 UNIX workstations. So when the CEO of your organization comes in to work on Monday, his secretary finds the following in his email

box:

To: John Smith, CEO From: Dark Haqr Subject: Rans0m

I 0wN U L^m3r It wi11 c0st u a kwart3r Mi11i0n t0 g3t ur dAtA b^k.

What is our SLE at this point? We could say $250,000, but it might not be quite that simple. If there were backups, we might be able to restore from backups and just lose a day or two of work. If there aren't backups (please, please ensure there are always backups), we have a

more interesting problem. At this point, we don't actually know if we will ever get the encryption key. The threat is that we will not. So, the value of the assets is the value of the data on these systems, plus the time to rebuild them from scratch, plus the loss from the downtime. How do we calculate the value of the data?

The value of data can be approximated by the burdened labor rate of the people who have been working on the system for the life of the project(s) on the system. To keep the numbers simple, we will consider each of the UNIX systems to be a professional's desktop. They are working on a single project that is two years along and they each make $60k, but their burdened rate (benefits, office space, and so on) is $100k. Ten people at $100k, for two years is $2 million dollars. What is our degree of exposure? It's 100 percent; the files are already encrypted. So, we quickly see that paying the quarter million and keeping our big mouths shut and not involving law enforcement is probably in our best interest. So in this scenario, we pay the money, get the key, and get back to work and everyone is happy. Now, what happens if we

don't fix the vulnerability?

Frequency of Threat—Annualized

Annualized loss expectancy (ALE) occurs when a threat/vulnerability pairing can reasonably be expected to be consummated more than once in a given year. In a brief given to the Joint Computer Security Conference in March 2000, Dr. Gene Schultz postulated this might be an inadequate measurement. Given the nuclear bomb example in our small town, this can't happen; indeed, we drop as many bombs as we want on the town, but we aren't likely to cause any further damage. ALEs fit very well into models such as shoplifting, returns in the mail-order business, and defaults on loans. In a competitive environment (e-business, for example), however, how many ALEs events can you survive? Consider the case of distributed denial of service. If your web storefront is shut down four or five times in a month, some of that business goes to your competition. How do you recover from that? How do ALEs factor into information assurance and intrusion detection?

I mentioned earlier that intrusion-detection technology is easily applied to unauthorized use detection. I also think that this can be a waste of skilled intrusion-detection analysts. But, there is a powerful business argument that says this is a very wise use of the system and personnel. As we work through the following example, note that even though I kept the numbers ridiculously low, we still ended up with some serious money, enough to pay the burdened rate of those entry-level professionals the organization says it can't afford. Use the following formula

to calculate ALE:

SLE x Annualized rate occurrence = Annual loss expectancy

This is nothing more than our SLE times the number of times it could be expected to occur in a year. This is why we ended the encrypted file system example with the question, "What happens if we don't fix the tooltalk vulnerability?" Dark Haqr takes our money, goes out and buys a Beamer, his friends inquire of the means of his sudden fortune, and we get to play the game again.

Let's do a common example: Web surfing on the job rather than working. First, we need to calculate an SLE. Say we have 1,000 employees, 25 percent of which waste an hour per week

surfing:

$50/hr x 250 = $12,500

To calculate the ALE we observe, they do it every week except when on vacation: $12,500 x 50 = $625,000

You can see why an organization might want to leverage its investment in intrusion-detection equipment and personnel to curb unauthorized use. Again, I kept the numbers much lower than what I have observed to be the case at many sites. Also, in the real world, the waste doesn't tend to be spread evenly across employees, but rather is localized in a small number of employees. If these employees can be identified and canned (after all, if they weren't working, they probably aren't really needed), there are a number of potential savings for the organization.

Recognition of Uncertainty

Соседние файлы в предмете [НЕСОРТИРОВАННОЕ]