Skip to main content
AdiFence

Product

A security-engineering approach, built in the open.

AdiFence is building behavior-first protection for ecommerce and cloud environments. This page describes the approach and how far it has come.

Developed and evaluated in a controlled lab environment. Not yet offered as a production service.

The approach

Four ideas that shape every control.

01

Behavioral observation

We record how sessions move through a shop — which routes they use, how often, in what rhythm and in which order — as telemetry tied to a pseudonymous session. The question is what a session does, not who it claims to be, and client addresses are hashed rather than stored.

02

Explainable detection

A detection is not a verdict. Each observation names the behavior it looked for, the values it measured, the conditions it checked (and whether each held) and a plain-language explanation. When there is too little evidence it says so instead of guessing.

03

Risk context

Observations are grouped into distinct behavioral families, and the result is shown together with how much of the evidence could actually be evaluated. A score is descriptive, not a probability, and a score of zero is not proof that a session is safe.

04

Route-aware protection

Browsing, stock checks, cart changes and checkout carry different risks and different costs of being wrong. A protection policy is therefore defined per kind of route, and a fast checkout is not treated as an abusive one.

Risk context

A number is never the whole answer.

Each behavioral family has an explicit state, so missing evidence is never silently turned into good news.

Observed
At least one supporting observation was seen.
Not observed
Everything relevant was evaluated and nothing was seen.
Partially evaluated
Some evidence could not be evaluated, so absence is not established.
Not evaluated
There was not enough evidence to evaluate.

Route-aware protection

Different journeys, different policies.

Browsing
Passive and frequent. Bursts from page loads and prefetching are normal, so allowances are generous.
Stock checks
Cheap to call and easy to automate. A small burst allowance and a slower sustained rate.
Cart changes
They change state, and rapid repetition rarely has an innocent explanation.
Checkout
High value. The policy is deliberately careful about false positives.

Observe before enforcement

A safe path from measuring to protecting.

A control earns the right to act. AdiFence moves a protection through five stages and does not skip any of them.

  1. 01

    Observe

    Record behavior and measure it before judging it.

  2. 02

    Explain

    Turn measurements into signals that state why they were raised.

  3. 03

    Dry-run

    Decide what a protection would do, and record it without acting.

    Where the lab is today

  4. 04

    Validate

    Compare the recorded decisions with reality and tune the policy.

  5. 05

    Enforce

    Only after validation, and only deliberately, apply a protection.

    Not active

Today the lab stops at dry-run: decisions are recorded, nothing is blocked, delayed or rejected.

Plainly

What this is not.

  • A production service. AdiFence is not yet offered as one.
  • Customer deployments, references or case studies.
  • Blocking rates, accuracy figures or traffic statistics.
  • Service-level commitments or guarantees of any kind.
  • A promise to stop every bot.

Contact

Let's build safer digital systems.

Working on a cloud environment, security platform or high-demand ecommerce experience?

Contact AdiFence