Shopping Is The Attack - A Decade Of E Commerce Scalper Wars, And The Multi Agent AI Era

The talk opens with a war story from a former head of security operations at a large European online retailer. A defensive block on a bag API, meant to stop sneaker bots, also took down an internal image service for about 45 minutes, while a limited release sold out within a minute. The speaker argues that scalping needs no vulnerability: each step is a valid request, so standard controls struggle to separate shoppers from bots.

The middle section walks through the attacker's playbook. It covers building an account pool through weakly validated student sign-ups and credential stuffing, then scoring accounts for loyalty so they pass checkout checks. Pre-release catalogue data and product images are pulled from unprotected APIs, computer vision matches images to SKUs, and a watcher polls public catalogues and alerts group chats the moment stock appears. The speaker also describes hoarding inventory to push prices down before buying, and how a raffle scandal and paid all-in-one bot tools lowered the barrier to entry. Rate limits, user agent rules, schema checks and reputation scoring each fell short, often because the business resisted added friction.

The closing part links this pattern to AI agents, which can run these steps in parallel at speeds beyond human operators. For practitioners, the takeaway is to treat shopping flows as an attack surface, audit mobile and internal APIs for exposed identifiers, and weigh collateral damage before blocking traffic.