Presets
Playback
Network
Size of the pool and how much of it is exposed.
Defences
Rate limiting, owner hardening and clean-up.
Scanning and time
Probe rate and how long the run lasts.
Actions
Camera
Quality
Layers
Model checks
The science
What this is (and is not)
An abstract epidemic model for defenders: how fast a compromise spreads through devices that still use factory-default credentials, and how much each defensive lever changes the outcome. It contains no scanning tool, no credential list and no exploit detail.
The model
Every device is in one of three states: S still on default credentials, I compromised, H hardened (unique credentials) or cleaned. A compromised device probes random addresses at rate s in an address space of Omega addresses; a probe reaches a given device with probability 1/Omega and gets through unless it is blocked by lockout or rate limiting (share lock). The hazard on a susceptible device is therefore a I with a = s (1 - lock) / Omega.
dS/dt = -a S I - r S, dI/dt = a S I - g I, dH/dt = r S + g I, where r is the rate at which owners change the default credentials and g the rate at which a compromised device is detected and cleaned. The page integrates this with RK4 and also runs a stochastic agent version (every device is its own random variable, advanced with the exact per-device hazard).
Threshold and the four levers
Early growth rate is a S0 - g: the outbreak grows only if a S0 > g, i.e. R0 = a S0 / g > 1. Each lever maps to one parameter: unique per-device passwords or forced change at setup shrink S0; lockout and rate limiting shrink a; monitoring raises g; awareness campaigns raise r. Graphs 2 and 3 sweep two of them and show the threshold behaviour.
Final-size relation
With no hardening (r = 0) the number of devices that escape satisfies S_inf = S0 exp(-(a/g)(N_t - S_inf)), with N_t = S0 + I0. With no cleaning either the model is logistic: I(t) = N_t / (1 + (S0/I0) e^(-a N_t t)).
The three checks
(1) RK4 against the logistic closed form, (2) RK4 against the final-size relation, (3) the mean of 40 stochastic runs against the ODE at the time of peak infection, with a 3 standard-error band.
Simplifications
Well-mixed random scanning in a fixed address space; one probe is one chance; all devices are identical; no network topology, no patching of the vulnerability itself, no reinfection of cleaned devices. Probe rates and address space are scaled so that a run fits in minutes - real values differ by orders of magnitude, but a S0 / g is what matters. The drawn probe lines are a cosmetic sample.
What to try
- Run the default: about 60 % of devices on defaults, a few bots, and the compromised count rises, peaks and falls as owners harden and monitoring cleans.
- Choose "Factory-unique passwords": the outbreak dies out (R0 below 1).
- "Just below the outbreak threshold" sits right at a S0 = g: tiny changes to the default share flip it from fizzle to outbreak (graph 2).
- Raise the lockout slider and watch graph 3: blocking probes is as good as hardening some of the devices.
- "Fast scanner, no defences": nearly everything falls; then use "Harden everyone now" mid-run to see the compromised count decay at rate g.