Building requirements is one of the first steps in the SDLC, where we define the goals and objectives of our future application. Usually, at this phase, we collect relevant stakeholders and start discussing their needs and expectations. We talk to people who will use the application, those who will manage it, and anyone else who might be affected by it to understand what they want the application to do and how it should work.
After reading this article, we will understand how to:
- Prepare Security Requirements
- Properly classify and rate threats
- Use a language of money in the risk assessment and prioritization
- Make sure, that defined security requirements will be implemented during other steps of SDLC
During requirements engineering, we need to separate functional and security requirements. While functional requirements show what an application should do, security requirements show what an application shouldn't do.
It's important to carefully consider building security requirements because a large part of the system's security depends on them. Good security requirement follows SMART principle:
| Name | Definition |
|---|---|
| Specific | Requirement shouldn't be complex and too broad, but exact and clear. |
| Measurable | There should be a clear way to test if this requirement was met or not. |
| Achievable | Developers should have a clear understanding of actions they need to take in order to meet the requirement. |
| Relevant | Requirement should ensure it addresses actual risks and provide meaningful protection, avoiding measures, that don't add values. |
| Time-bound | There should be a clear timeframe for implementation of this security requirement. |
Usually, stakeholders involved in security requirements engineering are:
- Developers
- Security experts
- Project Managers/Architects
All of them can participate in one of the ways to build security requirements: Security and Abuser stories.
Security Stories
As a [role], I want [security feature], so that [benefit]
Security stories are a special type of user stories, that focus on the security of the system.
You should look at the application from the perspective of users and stakeholders who need protection.
Examples:
- As a developer, I want to ensure that all passwords are stored with strong hashing algorithm so that even if the database is compromised, the passwords remain secure.
- As a system administrator, I want to log all access to sensitive data so that we can audit and identify any unauthorized access.
Abuser Stories
As a [bad guy], I want to [do something bad]
Abuser stories are the opposite of security stories. Here you need to think like an attacker, finding the ways you can exploit an application.
Examples:
- As an attacker, I want to perform a brute force attack on the login page to gain access to user accounts.
- As an attacker, I want to intercept data transmitted over the network to steal sensitive information.
So, security and abuser stories allow us to look at the application from both points of view: the user's and the attacker's.
It is a proactive approach, that provides a detailed and scenario-based understanding of security requirements.
Now we need a comprehensive way to ensure our critical assets and potential risks are managed.
For this, we can use the FAIR model:
Factor analysis of information risk (FAIR)
FAIR is a methodology, that helps to assess and manage informational risk in a financial terms.
It includes the following core steps:
- Threat and Critical Asset identification - defining valuable assets of the application and identifying related threats for them.
- Contact Frequency (СF) assessment - calculating how frequently the vulnerability interacts with critical asset
- Calculating Probability of Action (PoA) - finding the probability that asset would be attacked
- Threat Event Frequency (TEF) assessment - multiplication of CF and POA
- Vulnerability (Vuln) assessment - the probability that the attack on the asset would be successful
- Loss Event Frequency (LEF) assessment - multiplication of TEF x Vuln
- Defining Loss Magnitude (LM) - calculating Primary Losses (actual damage in a result of the attack) and Secondary Losses (reputation, litigation losses)
- Calculating Overall risk - multiplication of LEF x LM
Sounds a bit hard, right? As FAIR is just methodology, not framework, there are no concrete ways of how you should calculate risks.
But using simple Threat Modeling techniques, we can cover most of these steps:
Threat Modeling
Threat modeling allows us to identify and rate threats.
Identifying threats helps us to understand which security aspects are at risk, while rating ensures we prioritize them in the right way.
To properly identify threat we will use STRIDE framework:
| Threat | Description | Broken Principle |
|---|---|---|
| Spoofing | Pretending to be as someone or something else. | Authentication |
| Tampering | Unauthorized alteration or modification of data/communications. | Integrity |
| Repudiation | Denial by involved party of previously performed actions due to lack of auditing and logging. | Non-repudiation |
| Information disclosure | Unauthorized access or exposure of sensitive data. | Confidentiality |
| Denial of Service | Disruption of system's normal functioning. | Availability |
| Escalation of privilege | Allowing entity to do something, what it shouldn't have permission for. | Authorization |
As we see, it allows us to classify a threat in one or more of 6 categories, defining which security aspects are affected.
After the identification we can calculate the risk by using the DREAD framework:
| Name | Definition |
|---|---|
| Damage Potential | How big the potential damage would be? |
| Reproducibility | How easy is it to reproduce this attack? |
| Exploitability | How hard is it to use this vulnerability? |
| Affected Users | How many users would be affected? |
| Discoverability | How fast can an attacker discover the vulnerability? |
Each category in the model is scored from 0 to 10. The sum of the scores in all categories is total risk score. Maximum risk score is 50.
As an example let's create a STRIDE and DREAD analysis for SQL Injection:
| Threat | Description | Mitigation |
|---|---|---|
| Spoofing | Manipulation of SQL queries to bypass authentication mechanisms. | Use prepared statements and parameterized queries to prevent manipulation of authentication queries. |
| Tampering | Changing data within the database by injecting malicious SQL commands. | Use data versioning for tracking of changes and rollback if unauthorized modifications are detected. |
| Repudiation | Altering transaction records, making it difficult to verify malicious actions. | Enable Database Activity Monitor (DAM) for logging and auditing actions within the database. |
| Information Disclosure | Retrieving sensitive information from the database. | Encrypt all sensitive data and store all keys separately from the data. |
| Denial of Service | Executing costly SQL queries to disrupt database operations. | Implement rate limiting and monitor query performance to detect and mitigate abuse. |
| Escalation of Privilege | Injecting to get elevated privileges within the application or database. | Enforce Least Privilege principle and use role-based access control. |
According to STRIDE classification, we can easily calculate DREAD scores:
| DREAD | Score |
|---|---|
| Damage Potential | 9 |
| Reproducibility | 8 |
| Exploitability | 10 |
| Affected Users | 10 |
| Discoverability | 8 |
Thus, the total risk score for SQL Injection is 9 + 8 + 10 + 10 + 8 = 45.
After this, we can use Security & Abuser stories, STRIDE and DREAD frameworks to structure our approach with FAIR methodology:
› Threat and Critical Asset identification
Using Security & Abuser stories, we can find critical assets.
- As a developer, I want all user input to be validated to prevent SQL Injection in the database.
From this security story we see, that database is our critical asset.
To identify the influence vectors of the threat, we will use STRIDE.
› Contact Frequency (СF) assessment
This criteria fully depends on the functional requirements.
For instance, if our critical asset is the database and vulnerability relates to it, the frequency is actually how often will the user interact with the database.
› Calculating Probability of Action (PoA)
We can use Reproducibility and Exploitability scores from the DREAD framework.
For instance, SQL Injection's Reproducibility score is 9, and the Exploitability's score is 10.
Then our PoA would be (9 + 10) / 20 = 0.95
› Threat Event Frequency (TEF)
As mentioned before, TEF = CF x PoA.
For example, if there are 100 user-side interactions with the database, then for SQL Injection:
TEF = 100 x 0.95 = 95 threat events per day.
› Vulnerability (Vuln) assessment
For the Vulnerability assessment we can use the final DREAD score.
SQL Injection's DREAD score is 9 + 8 + 10 + 10 + 8 = 45/50, or 0.99.
› Loss Event Frequency (LEF)
LEF = TEF x Vuln
Then for our scenario, LEF = 95 x 0.99 = 94 loss events per day.
› Loss Magnitude (LM)
Loss Magnitude is calculated by summing potential primary and secondary losses. At this step we don't use any threat modeling approaches, cause it requires more specific analysis.
For instance, let's calculate the imaginary SQL Injection's Loss Magnitude:
Primary Losses:
The potential cost of stolen data: 50.000$
Cost of restoration works: 30.000$
System downtime: 10.000$
Total : 90.000$
Secondary Losses:
Legal and regulatory losses: 60.000$
Increased security costs: 20.000$
Total: 80.000$
Thus, Total Loss Magnitude is 90.000$ + 80.000$ = 170.000$
› Overall Risk
Potential Overall Risk = LEF(94) x LM(90.000) = 8.460.000$ per day
By using STRIDE, DREAD, Security & Abuser Stories, and FAIR, we learned how to develop strong security requirements.
The great thing about FAIR is that in the end, it translates these risks into financial terms, making it much easier for management to understand the importance of each security measure. This is especially helpful since it's often challenging to convey the significance of security risks to top executives.
Now after we have our security requirements and know their financial impacts, we can ensure we don't miss anything by using a Secure Requirements Traceability Matrix (SRTM).
Security Requirements Traceability Matrix (SRTM)
SRTM is a detailed document that links security requirements to their implementation and testing.
It makes sure that all security needs are handled during development, showing a clear path from the start to the final tests.
| Requirement ID | Description | Source | Priority | Control Reference | Implementation Reference | Testing Method | Status | Comments |
|---|---|---|---|---|---|---|---|---|
| Unique identifier for each security requirement. | Detailed description of the security requirement. | Origin of the requirement (threat model, regulatory, etc) | The importance of the requirement | Reference to design document that address requirement. | Links to the code or configuration that fulfills the requirement. | The method used to verify requirement (unit/integrity/penetration tests) | Current status of the requirement (started/completed/in progress/verified) | Additional notes or comments |
Let's imagine that after using the FAIR framework with Security & Abuser Stories, we identified the following security requirements:
- Implement 2FA to follow PCI-DSS
- Using input validation to prevent XSS Injection
- Logging access to sensitive assets
In this case, our matrix will look like this:
| Requirement ID | Description | Source | Priority | Control Reference | Implementation Reference | Testing Method | Status | Comments |
|---|---|---|---|---|---|---|---|---|
| SR-01 | Implement 2-FA | PCI-DSS | High | Data Flow Diagram DFD-01 | IMP-01 | Integration Test: IT-01 | Completed | SMS and Email OTPs being used |
| SR-02 | Validate inputs to prevent XSS injection | OWASP TOP 10 | High | Design Document DD-02 | IMP-02 | Static Analysis: SA-01 | Completed | OWASP Netryx used for input validation |
| SR-03 | Log all access to sensitive data | GDPR | Medium | Design Document DD-03 | IMP-03 | Penetration Testing: PT-01 | Not Started | SIEM tool integration planned |
For building requirements traceability matrix you will use such tools like YouTrack or Jira.
Summary
In this article, we learned how important it is to build security requirements early in the SDLC. By talking to stakeholders and using methods like Security & Abuser Stories, we can spot critical assets and potential threats from both user and attacker perspectives.
We used STRIDE to identify threats, DREAD to assess them, and FAIR methodology to look at these threats from all angles and translate their impact into financial terms, making it easier for management to understand their importance.
Finally, we talked about the Secure Requirements Traceability Matrix (SRTM), which helps us track security requirements from start to finish.
This ensures that nothing is missed and all security needs are properly addressed.
Finding and fixing security issues during the requirements phase can save millions of dollars later on. It’s much cheaper to address these problems early rather than after the application is built or at later SDLC steps.
OWASP is a non-profit foundation that envisions a world with no more insecure software. Our mission is to be the global open community that powers secure software through education, tools, and collaboration. We maintain hundreds of open source projects, run industry-leading educational and training conferences, and meet through over 250 chapters worldwide.
Top comments (2)
This is a great resource. I’m a fan of working with our devs by using the Threat Modeling methodology by Shostack and Associates. I’ll drop a link here: shostack.org/books/threat-modeling...
Definitely will check the book today, thank you for sharing!