← Systems Lab

Monte Carlo cyber-risk simulator

What happens to the risk of losing money if a bank can recover more diverted payments? I built this browser prototype to explore that question and two other common security scenarios.

Jump to the simulator

Systems Lab

What could this scenario cost in a year?

Each run plays out the same 12 months. Some runs have no losses. Others have several. Change an assumption to see how the costs shift.

Supplier invoice
New bank detailsChange requested for the next payment
Callback pending

Finance receives a false request to change a supplier’s bank details. A callback to the supplier or a timely bank recall could prevent a loss.

What counts as loss The amount of the transfer the bank cannot recover.

FBI · Business Email Compromise
Diverted supplier payment
Bank details are changed
Payment goes out before discovery
Some of the money stays missing

Each percentage is the chance of completing that step, if the preceding steps have already happened.

Chance one starting event leads to loss
—
Expected events per year (λ)
—
Annual chance of at least one event
—

Simulated average annual loss
—
Annual loss Value at Risk (VaR) 95%
—
Runs with a loss
—

Check the simulation against the formula

Possible losses over one year
€0—

View counts by loss range
Annual loss (€)Runs

Adjust the assumptions ↑

How the model works

Each run represents one possible outcome for one year. Choose a scenario, then set how often it starts and how likely each step is after the preceding steps. The model multiplies those probabilities by the starting rate to estimate how often a loss occurs.

The simulator draws the number of incidents in each run from a Poisson distribution. This assumes independent incidents at a steady rate. For each incident it draws a cost from the scenario’s triangular distribution, defined by the minimum, most likely and maximum cost. Adding the costs gives the total loss for that run. Some runs have no incidents and a loss of €0.

Reading the result

At 95% Value at Risk (VaR), at least 95% of the runs have a loss at or below the displayed amount. The threshold alone tells you little about the costliest outcomes. The histogram shows how far the sample extends above it.

Try lowering the chance that a file is opened before access is revoked. Keep the other inputs fixed and watch how the share of outcomes with a loss changes. The fixed random seed makes the comparison repeatable. The probabilities and costs are synthetic. To estimate the effect of a faster response in practice, you would need evidence such as the time between sharing, opening and revoking access.