MIME Parsing Checklist for Financial Services

Create finance-only MX subdomains (e.g., ap.inbound.example.com) and require TLS 1.2+ via MTA-STS, with reporting enabled to detect downgrade attempts.

Accept emails only from vendors whose From domain aligns with a passing DMARC policy, and keep an explicit allowlist for payees and banks to reduce spoofing risk.

Assign tagged addresses like ap+entity123@inbound.example.com and map tags to ledgers or client accounts to simplify downstream routing.

Permit only PDF, CSV, OFX/QFX, and XML (ISO 20022) and block executables, script files, and unusually large archives to reduce attack surface and processing load.

Specify cutoffs like T+1 02:00 UTC for statements and invoices, normalize timestamps to UTC, and alert if emails arrive outside the window.

Create separate inboxes and credentials for each business unit and environment (prod, staging) to prevent cross-tenant data exposure.

Maintain versioned contracts between inboxes and downstream ERPs or core banking systems so schema changes do not break processing.

Compute and record authentication outcomes on every message, quarantine failures, and expose results in audit fields for investigations.

Require TLS and block cleartext SMTP, publish and enforce MTA-STS, and alert on validation problems or downgrade attempts.

Manage private keys in HSM or cloud KMS, rotate them periodically, decrypt messages at ingest, and retain original encrypted blobs for audit.

Run multi-engine AV and sandboxing for PDFs with JavaScript, Office macros, and archives; reject or quarantine risky content.

Identify SSNs, PANs, routing and account numbers, and mask or tokenize them; store non-reversible hashes where identifiers are required for linking.

Canonicalize the parsed JSON, compute a SHA-256 digest, chain it to the prior event, and optionally time-stamp using RFC 3161 for non-repudiation.

Store required records in WORM where applicable and purge raw MIME after mandated retention windows, with automated, auditable deletion.

Separate roles for key management and parsing operations, require MFA for consoles, and apply IP allowlists for administrative endpoints.

Normalize all text parts to UTF-8 and respect declared charsets to prevent mangled vendor names and invoice fields.

Prefer text/plain when available, otherwise sanitize HTML to remove scripts and trackers before extraction.

Retain Message-ID, In-Reply-To, References, and a pruned Received chain; use them for idempotency and conversation context.

Resolve cid: links to embedded parts, extract inline images when relevant, and ignore known tracking pixels to reduce noise.

Support Outlook-generated TNEF to extract invoices or statements that might otherwise be lost.

Use a permissive parser that records RFC violations, quarantines severe cases, and never silently drops financial emails.

Capture validity, signer certificate subject, and chain status; require a valid signature for sensitive payment instruction mailboxes.

Store original filename, declared MIME type, detected type, size, positional index, and SHA-256 for deduplication and audit.

Capture vendor name, invoice number, line items, taxes, and totals; cross-check against subject and body text and raise exceptions on mismatches.

Auto-detect delimiter and charset, enforce expected column sets, and reject files with missing or unexpected headers.

Support camt and pain families, enforce schema version compatibility, and extract IBAN, BIC, and amounts with precise data types.

Extract FITID, posted date, and ledger balance; dedupe on FITID and normalized date-amount to prevent double posting.

Invoke OCR when no text layer exists and require a confidence threshold (e.g., 0.98) before releasing records to accounting systems.

Parse currency codes, store high-precision amounts, and retain original and converted values with the FX rate source and timestamp.

Detect replacements via subject tokens or filename versions and supersede older records instead of creating duplicates.

Compute SHA-256 for each attachment and skip reprocessing when the same document arrives in a later email or thread.

Sign each payload with a shared secret and include a timestamped header; require mTLS for enterprise consumers handling payments.

Derive an event ID from Message-ID plus attachment checksums and require consumers to treat duplicate deliveries as safe.

Use per-mailbox sequence numbers or Received-Date ordering and avoid parallel processing for the same thread or account.

Use exponential backoff with jitter, cap retries, and route to a dead-letter queue with reason codes and requeue tooling.

Provide since-id cursors, pagination, and ETags for consumers that cannot accept webhooks, and throttle politely on 429s.

Track parse latency, error rate by sender, document type mix, DMARC fail rate, and webhook success, with correlation IDs for traces.

Send fixture emails covering PDFs, CSVs, ISO 20022, and OFX, and assert golden JSON outputs to catch regressions early.

Store raw source in encrypted object storage with checksums and support replay into newer parser versions for audits and bug fixes.