fix(export): neutralize CSV/DDE formula injection in exported text

Exported journal notes and scraped event titles could begin with a formula
trigger (= + - @, tab, CR) that Excel/LibreOffice execute on open.
Csv.NeutralizeFormula apostrophe-prefixes such cells so they render as text;
applied to user notes, raw event ids and scraped titles. Numeric/date cells
the exporter formats itself stay numeric for downstream analysis.
This commit is contained in:
2026-05-29 13:57:18 +03:00
parent 88615a95e9
commit 08486667c3
4 changed files with 70 additions and 4 deletions
+15
View File
@@ -44,4 +44,19 @@ public static class Csv
return value; return value;
return "\"" + value.Replace("\"", "\"\"") + "\""; return "\"" + value.Replace("\"", "\"\"") + "\"";
} }
/// <summary>
/// Defuses spreadsheet formula / DDE injection: when a cell would start with a formula
/// trigger (<c>= + - @</c>, tab or CR) it is prefixed with an apostrophe so Excel /
/// LibreOffice render it as text. Apply to USER-supplied or SCRAPED text fields (notes,
/// event titles) before they enter a row — numeric/date cells your own code formats are
/// trusted and don't need it (keeping them numeric for analysis).
/// </summary>
public static string NeutralizeFormula(string? field)
{
var value = field ?? string.Empty;
return value.Length > 0 && value[0] is '=' or '+' or '-' or '@' or '\t' or '\r'
? "'" + value
: value;
}
} }
@@ -60,7 +60,7 @@ public sealed class ExportToCsvUseCase
{ {
b.PlacedAt.ToString("yyyy-MM-dd HH:mm", CultureInfo.InvariantCulture), b.PlacedAt.ToString("yyyy-MM-dd HH:mm", CultureInfo.InvariantCulture),
Title(titles, b.EventId), Title(titles, b.EventId),
b.EventId.Value, Csv.NeutralizeFormula(b.EventId.Value),
b.Selection.Type.ToString(), b.Selection.Type.ToString(),
b.Selection.Side.ToString(), b.Selection.Side.ToString(),
b.Selection.Value?.Value.ToString(CultureInfo.InvariantCulture) ?? string.Empty, b.Selection.Value?.Value.ToString(CultureInfo.InvariantCulture) ?? string.Empty,
@@ -68,7 +68,7 @@ public sealed class ExportToCsvUseCase
b.Stake.ToString(CultureInfo.InvariantCulture), b.Stake.ToString(CultureInfo.InvariantCulture),
b.Outcome.ToString(), b.Outcome.ToString(),
b.NetProfit?.ToString(CultureInfo.InvariantCulture) ?? string.Empty, b.NetProfit?.ToString(CultureInfo.InvariantCulture) ?? string.Empty,
b.Notes ?? string.Empty, Csv.NeutralizeFormula(b.Notes),
}); });
return await WriteAsync("journal", Csv.Document(header, rows), ct).ConfigureAwait(false); return await WriteAsync("journal", Csv.Document(header, rows), ct).ConfigureAwait(false);
@@ -93,7 +93,7 @@ public sealed class ExportToCsvUseCase
{ {
b.OpenedAt.ToString("yyyy-MM-dd HH:mm", CultureInfo.InvariantCulture), b.OpenedAt.ToString("yyyy-MM-dd HH:mm", CultureInfo.InvariantCulture),
Title(titles, b.EventId), Title(titles, b.EventId),
b.EventId.Value, Csv.NeutralizeFormula(b.EventId.Value),
b.PickedSide.ToString(), b.PickedSide.ToString(),
b.Rate.ToString(CultureInfo.InvariantCulture), b.Rate.ToString(CultureInfo.InvariantCulture),
b.Stake.ToString(CultureInfo.InvariantCulture), b.Stake.ToString(CultureInfo.InvariantCulture),
@@ -127,6 +127,8 @@ public sealed class ExportToCsvUseCase
return titles; return titles;
} }
// Titles are scraped ("Side1 vs Side2") so they're treated as untrusted text and
// neutralized against CSV/formula injection.
private static string Title(IReadOnlyDictionary<DomainEventId, string> titles, DomainEventId id) => private static string Title(IReadOnlyDictionary<DomainEventId, string> titles, DomainEventId id) =>
titles.TryGetValue(id, out var t) ? t : id.Value; Csv.NeutralizeFormula(titles.TryGetValue(id, out var t) ? t : id.Value);
} }
@@ -38,4 +38,32 @@ public sealed class CsvTests
"Kelly,ok\r\n" + "Kelly,ok\r\n" +
"\"Flat, fixed\",\"say \"\"hi\"\"\"\r\n"); "\"Flat, fixed\",\"say \"\"hi\"\"\"\r\n");
} }
[Theory]
[InlineData("=cmd|'/c calc'!A1")]
[InlineData("+1+1")]
[InlineData("-2+3")]
[InlineData("@SUM(A1)")]
[InlineData("\ttab")]
[InlineData("\rcr")]
public void NeutralizeFormula_PrefixesLeadingFormulaTriggers(string dangerous)
{
Csv.NeutralizeFormula(dangerous).Should().Be("'" + dangerous);
}
[Theory]
[InlineData("Home vs Away")]
[InlineData("normal note")]
[InlineData("3-1 win")]
[InlineData("")]
public void NeutralizeFormula_LeavesSafeValuesUntouched(string safe)
{
Csv.NeutralizeFormula(safe).Should().Be(safe);
}
[Fact]
public void NeutralizeFormula_Null_IsEmpty()
{
Csv.NeutralizeFormula(null).Should().BeEmpty();
}
} }
@@ -74,6 +74,27 @@ public sealed class ExportToCsvUseCaseTests : IDisposable
content.Should().Contain("evt-1"); content.Should().Contain("evt-1");
} }
[Fact]
public async Task ExportJournalAsync_NeutralizesFormulaInjection_InNotes()
{
var bet = new PlacedBet(
Guid.NewGuid(),
new EventId("evt-1"),
new Bet(MatchScope.Instance, BetType.Win, Side.Side1, null, new OddsRate(2.0m)),
100m, T0, BetOutcome.Pending, "=cmd|'/c calc'!A1");
_bets.ListAsync(Arg.Any<CancellationToken>()).Returns(new[] { bet });
_events.GetManyAsync(Arg.Any<IReadOnlyCollection<EventId>>(), Arg.Any<CancellationToken>())
.Returns(new Dictionary<EventId, Event>());
var path = await CreateSut().ExportJournalAsync();
var content = await File.ReadAllTextAsync(path!);
// The dangerous note is apostrophe-prefixed so no cell opens as a live formula in
// Excel; it sits in the last column, preceded by the CSV separator.
content.Should().Contain(",'=cmd");
content.Should().NotContain(",=cmd");
}
[Fact] [Fact]
public async Task ExportPaperLedgerAsync_ReturnsNull_When_NoBets() public async Task ExportPaperLedgerAsync_ReturnsNull_When_NoBets()
{ {