web security audit fixes

web/src/server/services/voiceprint/audio.processor.ts

@@ -41,6 +41,11 @@ interface WavHeader {
 
 /** Maximum allowed audio duration in seconds */
 const MAX_DURATION_SEC = 30;
+/** Maximum raw WAV file size before processing (default 5MB). Prevents memory exhaustion. */
+const MAX_INPUT_BYTES = parseInt(
+  process.env.VOICEPRINT_MAX_INPUT_BYTES ?? "5242880",
+  10,
+);
 /** Target normalization level in dBFS */
 const TARGET_DBFS = -3;
 /** Frame size for VAD in milliseconds */
@@ -359,16 +364,27 @@ function computeQualityMetrics(samples: Float64Array): {
 
 /**
  * Main audio preprocessing pipeline:
- * 1. Parse WAV header
- * 2. Read PCM samples
- * 3. Convert to mono
- * 4. Resample to 16kHz
- * 5. Normalize to -3 dBFS
- * 6. VAD silence trimming
- * 7. Limit to 30 seconds
- * 8. Convert to 16-bit PCM
+ * 1. Validate input size
+ * 2. Parse WAV header
+ * 3. Validate duration from header (reject too-long audio before decoding)
+ * 4. Read PCM samples
+ * 5. Convert to mono
+ * 6. Resample to 16kHz
+ * 7. Normalize to -3 dBFS
+ * 8. VAD silence trimming
+ * 9. Limit to 30 seconds
+ * 10. Convert to 16-bit PCM
  */
 export async function preprocessAudio(inputBuffer: Buffer): Promise<ProcessedAudio> {
+  // Reject oversized input early to prevent memory exhaustion
+  if (inputBuffer.length > MAX_INPUT_BYTES) {
+    throw new Error(
+      `Audio file too large: ${(inputBuffer.length / 1024 / 1024).toFixed(1)}MB. ` +
+      `Maximum ${(MAX_INPUT_BYTES / 1024 / 1024).toFixed(0)}MB. ` +
+      `Please upload a shorter audio clip (max ${MAX_DURATION_SEC} seconds).`,
+    );
+  }
+
   // Detect if it's a WAV by checking RIFF header
   const isWav =
     inputBuffer.length >= 4 &&
@@ -382,6 +398,17 @@ export async function preprocessAudio(inputBuffer: Buffer): Promise<ProcessedAud
   }
 
   const { header, dataOffset } = parseWavHeader(inputBuffer);
+
+  // Validate duration from header BEFORE allocating sample buffers.
+  // This prevents loading multi-hour WAV files into memory.
+  const totalSamples = Math.floor(header.dataSize / (header.bitsPerSample / 8) / header.numChannels);
+  const durationSec = totalSamples / header.sampleRate;
+  if (durationSec > MAX_DURATION_SEC + 30) {
+    throw new Error(
+      `Audio too long: ${durationSec.toFixed(1)}s. Maximum ${MAX_DURATION_SEC}s for analysis. ` +
+      `Please trim your audio before uploading.`,
+    );
+  }
   let samples = readPcmSamples(inputBuffer, header, dataOffset);
 
   // Convert to mono