DLP Alert Triage

Most DLP alerts are the business doing its job — payroll emailing payroll data, sales
sending a contract. Work the alert on the evidence: what data, what destination, what
pattern — reaching a verdict without turning the desk into a surveillance operation on the
client's employees. Work it in order:

1. Extract the alert anatomy: policy/rule that matched, data classification and match
   count (e.g. "42 payment-card patterns"), the actor (user), the vector (email, cloud
   share, USB, print), the destination (recipient domain, share target, device), and the
   platform action (blocked, warned, logged-only). Route to the correct client per
   security-alert-response routing if on shared intake.
2. Business-context check — the false-positive test: does this flow match the actor's role
   and a routine process? Look up the actor's role and search for the same rule + same flow
   history: a documented recurring benign pattern (same sender, same destination, same
   cadence) points to false positive — confirm it still applies, never auto-close on
   history alone.
3. Exfiltration-signal check — what makes it real: destination (personal email/cloud, a
   competitor's domain vs an established counterparty); volume/spread (bulk or unusually
   broad for that role); timing/cluster (off-hours, or paired with a resignation notice,
   access-revocation ticket in flight, recent account alerts); behavior around controls
   (retrying after a block, renaming/zipping to evade patterns). None alone is a verdict;
   accumulation raises the tier.
4. Privacy-respectful investigation discipline: examine only what the alert and verdict
   require — metadata (what pattern, where to, how much) before content; open actual
   content only when the verdict genuinely requires it AND the client's policy permits desk
   access, and record that it was accessed and why. Investigate the event, not the person's
   mailbox at large. No verdicts about intent.
5. Verdict and path:
   - Business-process false positive → close with the evidence and feed recurring ones to
     security-noise-tuning for a NARROW policy exception (this sender-group to this
     destination — not "disable the rule").
   - Real exfiltration signals, or resignation/offboarding context → do NOT confront the
     user and do not investigate further solo: preserve the evidence and hand to
     insider-risk-basics for the escalation-to-client-leadership path. Odd hours + other
     account anomalies → possible compromise, security-alert-response /
     compromised-account-containment path.
   - Blocked-and-benign where the user just needs the legitimate path → answer the user's
     need so the block doesn't teach workarounds.
6. Document the decision, not just the action: anatomy, business-context findings, signals
   weighed, what content (if any) was accessed and under what authority, verdict, and
   route — plain-text note; classify per soc-classification-tree.

Guardrails — always:
- Employee privacy is a working constraint: metadata before content, content only with
  policy authority and a recorded reason, and never browsing beyond the event under
  investigation.
- No intent language in notes — "transferred 300 files to a personal account" is evidence;
  "trying to steal data" is a conclusion that belongs to the client's process.
- Real-signal cases stop here and go to insider-risk-basics — no solo deep-dives, no user
  confrontation, no tipping anyone off.
- Recurring false positives get tuned narrowly by the DLP platform owner; never
  blanket-disable a rule, never PSA auto-close.
- A capped or partial alert history is reported as such; verdicts name their evidence.
- Degradation: without DLP-console access, work from the alert body and direct the
  technician on what to pull; record what could not be verified. Never invent data.