What we hear
When fraud and bot defense owners come to us, they say:
Our on-sale went in under a minute and the seats were on resale within the hour. The vendor dashboard says it blocked two million bots. Nobody can tell me what got through.
The next drop is already on the calendar and my VP wants to know what we are changing. The vendor’s answer is a deck about what it caught.
We looked at having our own engineers replay it, but they do not have the tooling resellers use. Switching vendors takes a quarter we do not have.
What InnerLock does
Bot defense red team for ticket and product drops.
Not another detection layer to deploy, and not a vendor migration. An authorized adversary who tests what you already run, on your calendar.
-
Attack
With your written authorization, we run the resellers’ own playbook against your live site around a real or simulated launch.
-
Report
One page your CEO can read, and reproducible evidence your vendor cannot dispute.
-
Fix list
Prioritized by what can actually change before the next date on the calendar.
-
Retest
The same attacks again after the fix, before the launch. The next drop holds, and you know why.
What we attack with
Why we can
We know where these defenses look, and where they do not.
We spent three and a half years inside F5’s Shape bot defense team, reverse-engineering the five commercial bypass tools resellers were using and shipping a defense against each one. That is where the map comes from, and we have no vendor contract to protect.
- We know which signals they lean on. And which ones a current agent stack no longer trips. That is the difference between a test that looks busy and one that gets through.
- No vendor contract to protect. We do not resell your bot vendor and we are not bidding to replace it. The report says what got through, whoever that embarrasses.
- Evidence, not inference. Every named gap comes with the session that proves it and steps to reproduce it.
- Fixes you can apply where you already work. Changes land in the vendor, gateway, or WAF configuration you have, not in a product you have to buy from us.
- It fits between drops. A test window and a report sized to the weeks you actually have, not to a quarter-long migration.
See it in ten minutes
One of our agents completes a checkout on a site with a major bot vendor switched on.
Left: the agent’s session, step by step. Right: the vendor dashboard, still green. We run it live on a call, against a site we are authorized to test, and you ask whatever you want while it happens.
The side-by-side recording goes up after the first authorized test. Until then it runs live on a call — we would rather show you the real thing than a reconstruction.
Rules of engagement
An authorized adversary, with the paperwork to match.
We are asking you to let us attack your storefront. That only works if the boundaries are written down before anything runs.
- Written authorization first. Named scope, named targets, and an agreed test window, signed before the first request.
- Sized not to hurt you. Load and inventory impact agreed in advance, with a stop signal that works during a live launch.
- Real purchases handled by agreement. Test payment paths, order cancellation, and inventory restoration are settled in the scope, not improvised.
- Your data stays yours. Findings and session evidence are held under the agreement’s scope and retention terms, and customer secrets are never sent to third-party AI services.
- Reproducible, then retired. You get steps to reproduce every finding, and the tooling we built for your test is not reused against you.
Who we are
The person testing your defenses used to build them.
InnerLock is a San Jose company founded by Xiaofo (Shirley) Luo. Before InnerLock she spent about three and a half years on F5’s Shape bot defense team, on the side of the problem most teams only ever see through a dashboard.
- 5commercial bypass tools reverse-engineered on F5’s Shape team, with a shipped defense against each one
- 3.5years working bot defense at F5, inside the detection stacks retailers and ticketing platforms buy
That work is why the testing is adversarial rather than theatrical. Knowing how a detection stack decides what to trust is what tells you which agent setup will walk past it — and which change actually closes the door.
It is also why the report is written for the person who owns bot and fraud, not for a procurement committee: one named gap, the evidence behind it, and the configuration change that fixes it.
This is where the first customer’s account of a launch goes — in their words, with their name on it. We would rather leave it empty than fill it with someone who does not exist.
Contact
Running a drop or on-sale in the next 60 days?
Tell us the date and what happened at the last one, and we will say whether there is time for an authorized test before it — honestly, including if there is not. Security issues and responsible disclosure reach the same address.