BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//sf26-agenda//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:SESS-152@sf26-agenda
DTSTAMP:20261020T151709Z
DTSTART:20261105T183000Z
DTEND:20261105T191500Z
SUMMARY:Masking PII/PCI in Motion: Real-Time Data-Loss Prevention with Kafk
 a and Flink on Confluent Cloud
LOCATION:Room 2000
DESCRIPTION:Leo Bombonato\, Erick Da Silva\n\nMost data-loss prevention hap
 pens after the fact — batch scans\, nightly jobs\, audits that find the 
 leak days later. But sensitive data leaks in motion: an SSN typed into a f
 ree-text "notes" field rides a Kafka topic to a dozen downstream consumers
  within seconds. This session shows how we shifted DLP left\, into the str
 eam itself\, so PII and PCI are detected and masked before a record ever l
 ands in a downstream system — a pattern we took from lab to a real clien
 t engagement. We walk through a production-shaped pipeline: customer recor
 ds flow into Confluent Cloud Kafka; Confluent Cloud for Apache Flink runs 
 streaming SQL that calls a custom Java UDF; the UDF calls an external DLP 
 API (Nightfall) to find SSNs\, credit cards\, and emails hidden in free te
 xt. High-confidence findings are masked inline (last four digits kept) and
  attached as a structured findings object; lower-confidence hits are route
 d to a separate review topic for a human analyst instead of being silently
  dropped or over-redacted. Per-call latency is tracked per record and aggr
 egated per minute\, so the cost of "scanning in the stream" is observable\
 , not a black box. On technical depth\, we cover calling an external API f
 rom inside a Flink UDF (Confluent's External Connectivity for UDFs)\, keep
 ing secrets in a managed connection\, a fail-open design so a DLP outage n
 ever crashes the stream\, structured ROW return types matched to a JSON Sc
 hema sink\, and the whole pipeline provisioned as code with Terraform and 
 a single Makefile. Audience takeaways: a concrete pattern for in-stream re
 daction and tokenization\, how to balance automated masking against human 
 review to control false positives\, how to budget the latency and cost of 
 an external call per event\, and an honest account of what broke when we s
 hipped it to a client. By attending this session\, your contact informatio
 n may be shared with the sponsor for relevant follow up for this event onl
 y.
URL:https://sf26.rmoff.net/s/masking-pii-pci-in-motion-real-time-data-loss-
 prevention-152
END:VEVENT
END:VCALENDAR