9.5 Incorporating Probability into the Risk Analysis
So far this chapter has focused on an analysis technique based on scenario planning. We incorporated this technique in OCTAVE, because the lack of objective data for certain types of information security threats makes it difficult to incorporate a forecasting approach based on probability. However, we have found that there is considerable interest in using probability during a more traditional risk analysis. This section presents some basic concepts of probability and shows how you can include probability in the activities of process 7.
9.5.1 What Is Probability?
We define probability as the likelihood that an event will occur. We first consider the classical concept of probability. This concept is the oldest historically and was originally developed in connection with games of chance [Bernstein 96]. For example, consider a die, which is simply a cube with six faces. Because of its symmetry, each face is as likely to come up as any other. Thus, you could easily determine the probability of one face coming up with a roll of the die as 1 in 6. The key for this concept of probability is that all possibilities must be equally likely to occur.
Next, we consider the frequency interpretation of probability. This interpretation indicates that the probability of an event occurring (or a given outcome occurring) is the proportion of the time that similar events will occur over a long period of time. Note that when using the frequency interpretation of probability, you cannot guarantee what will happen on any particular occasion. Thus, you actually never really "know" the probability that an event will occur, because you will not be able to collect enough information to know precisely what will happen in the long run. Although you cannot know the exact value of a probability, you can estimate it by observing how often similar events have occurred in the past. Estimates of probability made after observing similar events are useful because of the law of large numbers [Freund 93]. This law states that as the number of times a situation is repeated becomes larger, the proportion of successes tends toward the actual probability of success. For example, consider multiple flips of a coin. If you flip a coin repeatedly and chart the accumulated proportion of time that you get heads, you will find that over time the proportion comes closer and closer to 1 in 2 (the probability of getting heads with each flip).
A common example that uses the frequency interpretation of probability is weather forecasting. If the forecast calls for a 60 percent chance of rain, it means that under the same weather conditions, it will rain in 60 percent of cases. Next, let's consider a variation of this casehow do you estimate the probability of something that occurs just once? Consider how doctors estimate the probability of how long it will take a patient to recover from an illness. A doctor can check medical records and discover that in the past, 50 percent of the patients recovered within two months under a specific treatment plan. By using this information from similar cases, the doctor can predict that there is a 50 percent probability that the patient will recover within two months.
You can probably see how complicated this can get. It is not always easy or straightforward to determine which cases are similar to the one that you are considering. In the case of the patient, the doctor might consider not just the treatment plan being prescribed but also the patient's age, gender, height, and weight, among other factors. This approach can be difficult and requires individual judgment, indicating how easy it is for two individuals to arrive at different probabilities for the same event.
The final type of probability that we will discuss is subjective probability. This approach is often used in situations where there is very little direct evidence. You might have only some indirect, or collateral, information, educated guesses, intuition, or other subjective factors to consider [Freund 93]. A person determines a probability based on what he or she believes to be the likelihood of occurrence. The key word here is "believes." Different people assess probabilities differently, based on their personal evaluation of a situation. One disadvantage of this approach is that it is often hard for people to estimate probability, and the same person can end up estimating different probabilities for the same event using different estimating techniques.
Probability and Information Security
In information security, you are interested in estimating the likelihood that a threat will actually materialize. For some types of security threats, you have information upon which you can draw. For example, you can use the frequency data to estimate the probability of natural disasters (e.g., floods, earthquakes) in your region. You might also be able to use the frequency of occurrence to estimate the probability of some systems problems, such as system crashes and susceptibility to viruses. However, for some other types of threats there are no frequency data.
How would you estimate the probability of an attacker viewing confidential customer data from your organization's customer database? How much company data do you have to estimate the probability of this attack? Most likely, your organization has not collected sufficient data about such attacks to enable an estimation of probability based on frequency of occurrence. If it has occurred, it has probably happened only once or twice. In addition, you cannot be sure how many times this attack has occurred but gone undetected. What about industry data? Is this the kind of information that companies readily disclose? Many attacks of this type go unreported, making it difficult to obtain sufficient data to derive probability based on frequency models. Finally, even if you had some industry data about these types of attacks, how do you establish which events are similar? For example, does information about past attacks in the banking sector apply to organizations in the manufacturing sector? All of these factors make a frequency-based estimation of probability difficult and time-consuming, if not impossible. That leaves us with subjective probability for threats resulting from human attackers.
Subjectively estimating probability for attacks by human threat actors is tricky. You need to consider the following factors:
Motivehow motivated is the attacker? Is the attacker motivated by political concerns? Is the attacker a disgruntled employee? Is an asset an especially attractive target for attackers?
Meanswhich attacks can affect your critical assets? How sophisticated are the attacks? Do likely attackers have the skills to execute the attacks?
Opportunityhow vulnerable is your computing infrastructure? How vulnerable are specific critical assets? (Note that this question is linked to the vulnerability data that you gathered in process 6.)
When estimating the above factors, people typically rely upon their experience to make educated guesses about the likelihood of attacks occurring. You would need experience with networked systems security as well as an understanding of the industry sector in which an organization operates. Note that some people do not have sufficient experience to estimate probability using subjective techniques. In fact, probabilities estimated by inexperienced people can actually skew the results of a risk analysis.
In general, you must be careful when incorporating probability into your risk analysis. The next section explains how you can incorporate probability into the activities of process 7 using a combination of frequency data and subjective estimation.
9.5.2 Probability in the OCTAVE Method
We propose using a combination of frequency and subjective probability into the OCTAVE Method's risk analysis activities. There are three activities to add if you choose to do this:
Describe the probability of threats to critical assets.
Create probability evaluation criteria.
Evaluate the probability of threats to critical assets.
Step 1: Describe the Probability of Threats to Critical Assets
In addition to identifying the impacts of threats, you identify probability. You gather information related to the factors that contribute to determining probability. Consider the following questions for each threat profile:
Which critical assets are likely targets of human threat actors?
What are the motive, means, and opportunity of each human threat actor who might use network access to violate the security requirements of the critical asset?
What are the motive, means, and opportunity of each human threat actor who might use physical access to violate the security requirements of the critical asset?
What historical data for your company or domain are available for all threats in the threat profile? How often have threats of each type occurred in the past?
What unusual current conditions or circumstances might affect the probability of the threats in the threat profile?
By answering the above questions, you gather both subjective and objective data about threats to your critical assets. You can then use them to estimate threat probability. Notice that the first three questions and the last question are subjective in nature, while the fourth question relates to any objective threat data you may have. You need to make sure that you record all subjective information and objective data for each type of threat to your critical assets.
Step 2: Create Probability Evaluation Criteria
In addition to developing evaluation criteria for impact, you also create evaluation criteria for probability. These criteria are measures against which you will evaluate each threat to establish a qualitative probability value for that threat. Evaluation criteria for probability indicate how often threats occur over a common period of time. When you create evaluation criteria for probability, you define measures for high, medium, and low likelihood of occurrence for your organization.
Review the probability information that that you gathered during the previous activity and answer the following questions:
What defines a "high" likelihood of occurrence?
What defines a "medium" likelihood of occurrence?
What defines a "low" likelihood of occurrence?
Remember, your goal is to define probability measures using any objective data that you have in addition to your subjective experience and expertise. You also need to make sure that your criteria are meaningful to your organization. As always, record your results.
Let's examine what evaluation criteria might look like for our sample organization. At MedSite the analysis team supplemented their skills by including the following:
A member of ABC Systems who had extensive information technology security experience. This individual understands the range of possible attacks in the medical domain and the degree of skill required to execute each attack.
A member from MedSite's risk management department. This individual has background knowledge about many of the threat actors in the threat profile.
The expanded team reviewed background information. The people with information technology and risk management expertise provided valuable insight into creating frequency ranges for each probability level. Figure 9-5 shows the resulting probability evaluation criteria.
Notice that the criteria in Figure 9-5 use frequency of occurrence to define probability levels. Team members used the data that they had for certain sources of threat in conjunction with their subjective experience for those sources for which they had little or no objective data. Thus, despite the use of frequency in the criteria, this represents a highly subjective look at probability, and it should be noted as such.
FIGURE 9-5 MedSite's Probability Evaluation Criteria
Step 3: Evaluate the Probability of Threats to Critical Assets
Finally, in addition to evaluating the impact of each threat, you evaluate its probability. Review all relevant background information before you complete this activity. Make sure that you review threat profiles for each critical asset and the evaluation criteria for probability.
Select a critical asset. Assign each threat a qualitative probability value (high, medium, or low) based on (1) the probability information that you have gathered, (2) the probability evaluation criteria that you created, and (3) your team's collective experience and expertise.
If you find that your probabilities don't make intuitive sensefor example, if all of your threats are evaluated as "high probability"you might want to go back and adjust your probability criteria. Once you are satisfied with your evaluation results, the final step is to add probabilities to the risk profile.
At MedSite the expanded team (the analysis team plus supplemental personnel) assigned probability values to each threat in all threat profiles. Figure 9-6 shows part of the PIDS risk profile with probability added to the tree.
FIGURE 9-6 Part of the PIDS Risk Profile (Including Probability): Human Actors Using Network Access Tree
This concludes process 7. Chapter 10, which examines risk mitigation, revisits the topic of probability and looks at building risk mitigation plans for each critical asset and forming a protection strategy for organizational improvement.