Skip to main content

The GDS Way and its content is intended for internal use by the GDS and CO CDIO communities.

Threat Modelling

What’s a threat?

A threat is an action that causes denial, alteration or disclosure of your assets that you are trying to protect. It could be a user database, a deployment pipeline or the integrity of web form submissions.

What’s threat modelling?

Threat modelling is analysing a service to raise potential information security and privacy issues.

The best time to perform threat modelling activities is during the design phase; however, it can be done anytime and should become a continuous process in your service team.

Within the Cabinet Office and GDS, we follow the Threat Modeling Manifesto’s four questions:

  1. What are we working on?
  2. What can go wrong?
  3. What are we going to do about it?
  4. Did we do a good enough job?

We answer most of the above over three sessions, with time between sessions allowing for asynchronous input.

1. Threat Discovery

Threat discovery aims to answer the “What are we working on?” and part of the “What can go wrong?” question.

We use an interactive exercise to talk through and explain the system or architecture. Someone who is comfortable with explaining the system from the service team will need to lead this part. Try to keep the scope small; it may be more effective to break a service down and perform multiple threat modelling activities.

The group of people involved typically includes the development team, cyber security and information assurance colleagues; however, working with other people, such as user researchers and product managers, often provides more varied perspectives.

As part of the interactive exercise, we surface hypothetical threats or vulnerabilities in a service. We use STRIDE to help us look at different aspects and make sure we get a rounded view of all threats.

1.1 Interactive Exercise

The purpose of this exercise is to talk through the service, architecture or the user journey with the people in the session, allowing for questions and answers.

Remotely, digital whiteboard software, such as Jamboard or Mural, can be used to add virtual post-its on top of the digital diagrams. Make sure the tools are accessible so all colleagues can participate and ensure access control is sufficient as the diagrams will likely be at least OFFICIAL-SENSITIVE when complete.

In person, which may be more appropriate for some teams and systems, diagrams are typically drawn up on a whiteboard, followed by questions and post-its of potential threats stuck at relevant points.

2. Threat Analysis

Threat analysis aims to finalise the answer to the “What can go wrong?” question. We use a scoring methodology to determine if a threat is valid and prioritise threats against each other.

2.1 Scoring

After the discovery stage, you can make a copy of the Threat Modelling Scoring template to use for noting down threats, the most likely actor, scoring, and tracking mitigation.

Scoring is useful as it helps prioritise what’s valid and what you should look into first.

The difficulty is a rating between 0 (very easy) to 5 (hard); it should take into account how hard it would be to gain access and any costs associated with performing the attack.

The reward is a rating between 0 (low) to 5 (high); it’s how valuable the outcome is to the threat actor.

The final score equals reward minus difficulty, so a high score is worse than a low one.

3. Threat Response

Threat response aims to answer the “What are we going to do about it?” question.

It may be helpful to add the outputs to your bug or ticket tracking system.

There are three typical responses to threats:

3.1 Mitigate

Implementing a preventive control, such as an architectural change, eliminates the vulnerability or blocks the threat.

3.2 Monitor

Monitoring is a reactive control to detect if a threat occurs; it controls a situation and limits damage or loss by detecting an attack early.

3.3 Acknowledge the risk

Acknowledging the risk but not planning any immediate action can help with resourcing any future mitigation or deciding to accept a risk because the cost of the mitigation outweighs the maximum cost of a threat.

4. Did we do a good enough job?

After completing the threat modelling activity sessions, you should set aside some time to review the threat modelling outputs regularly. This is to see whether you are improving the security and taking the mitigations forward; it’s also important to consider any new threats.

A recommendation is to perform this review within your service team every two months; the time you need may vary depending on the scope of your threat modelling, but 90 minutes is usually sufficient.

You should consider running the threat modelling sessions every year or when there are significant architectural changes planned.

Why’s it necessary?

Adam Shostack has produced an informative short video on this: Why Threat Model?

Adam explains that we threat model to anticipate problems when it’s inexpensive to deal with them.

How do you know what a problem is? Information and computer security is often broken up into three main pillars, known as the CIA triad:

Confidentiality

  • Information accessible to authorised people

Integrity

  • Information that’s correct

Availability

  • Information available when needed

Abuse or neglect of one or more of these pillars could lead to a security incident.

The opposite of the CIA triad is sometimes known as the DAD triad:

  • If confidentiality is affected, disclosure occurs.
  • If integrity is affected, alteration occurs.
  • If availability is affected, denial occurs.

Here is a fun, alternative look at the CIA and DAD triads, explained with Lord of the Rings references.

STRIDE

STRIDE is a model of threats developed by Microsoft. It’s a mnemonic for splitting security threats into six categories.

It’s a helpful tool for thinking about different potential attacks or vulnerabilities.

Threat Security Control
Spoofing Authenticity
Tampering Integrity
Repudiation Non-repudiability
Information Disclosure Confidentiality
Denial of Service Availability
Elevation of Privilege Authorisation

Spoofing

Spoofing occurs where malicious actors use another person’s user credentials without their knowledge.

It also includes attacks where systems mimic or disguise themselves as other systems, such as IP address spoofing in DNS amplification attacks.

The desired property is authenticity, meaning an event or action is trustworthy and genuine.

Tampering

Only authorised users should be able to modify a system or the data it uses.

The desired property is integrity, meaning an event or action is valid and correct.

Repudiation

Repudiation means an action or event cannot be associated with what triggered it, whether that is a person or another system component.

The desired property is non-repudiation, which means an action or event is difficult to deny by a person or system.

Information Disclosure

Data such as passwords or personal information is made available to those who should not have access to it.

The desired property is confidentiality, where data is only available to those who are authorised to view or alter it.

Denial of Service (DoS)

Data is unavailable for its intended audience.

The desired property is availability, where data is always available when those authorised to view or alter it expect it to be.

Distributed Denial of Service (DDoS) is a form of DoS where multiple entities are responsible for inappropriate traffic leading to overwhelmed services.

Elevation of Privilege

Data is available to unintended or unauthenticated users.

The desired property is authorisation, where data is only available to those expected to view or alter it.

This page was last reviewed on 25 June 2021. It needs to be reviewed again on 25 December 2021 by the page owner #gds-way .
This page was set to be reviewed before 25 December 2021 by the page owner #gds-way. This might mean the content is out of date.