BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Open-source Framework for Synthetic Fraud Datasets via Generative 
 AI - Insights from Industrial Secondment at IBM France
DTSTART:20250716T120000Z
DTEND:20250716T130000Z
DTSTAMP:20260719T055800Z
UID:indico-event-3920@indico.e5.physik.tu-dortmund.de
CONTACT:maik.becker@tu-dortmund.de\;serena.maccolini@cern.ch
DESCRIPTION:Speakers: Micol Olocco (TU Dortmund)\n\nThis work presents a f
 lexible and open-source framework for simulating realistic banking transac
 tion data\, designed to address the dual challenges of data scarcity and p
 rivacy compliance in financial fraud research. Developed during an industr
 y secondment at IBM France Lab Saclay\, within the European SMARTHEP Netwo
 rk\, our solution leverages Generative AI Agents to mimic both legitimate 
 and fraudulent behaviors in temporal transaction sequences.\nInitially bas
 ed on Markov models\, the framework has evolved into a Large Language Mode
 l (LLM)-driven architecture. A dedicated Strategist LLM defines behavioral
  blueprints for diverse profiles—such as everyday spenders\, travelers\,
  identity thieves\, or money launderers. Then\, an Agent LLM generates det
 ailed\, structured activities including transaction type\, timestamp\, loc
 ation\, amount\, and contextual signals such as device or network anomalie
 s. This two-stage design allows for flexible scenario generation and impro
 ved interpretability. To address limitations in output validity and format
 ting\, we explored the use of lightweight LLMs for output validation and c
 orrection. This self-correcting mechanism aims to reduce invalid generatio
 ns and enhance consistency.\nThe resulting synthetic datasets could offer 
 a safe and configurable environment for developing and testing fraud detec
 tion strategies—without the constraints of proprietary banking data. In 
 addition\, exploratory thinking on applying automation principles and LLMs
  to LHC research and infrastructure will be presented\, along with an over
 view of current community efforts.\n\nhttps://indico.e5.physik.tu-dortmund
 .de/event/3920/
LOCATION:CP-03-123 (TU Dortmund)
URL:https://indico.e5.physik.tu-dortmund.de/event/3920/
END:VEVENT
END:VCALENDAR