fix(formats): preserve high bits in remaining magic-byte checks (#2384)

* fix(formats): preserve high bits in remaining magic-byte checks

Node's "ascii" decoder masks the high bit, so byte-for-byte signature
comparisons accepted lookalikes such as CD DA ("MZ" | 0x80) or a
high-bit "bplist00"/"NIBArchive" header. ac57c88ef fixed this for the
shared DOS/MZ reader; apply the same latin1 decoding to the remaining
signature checks: NativeAOT and Windows native loader MZ/PE, binary
plist and NIBArchive detection, and PNG IHDR checks.

Display-name decodes (section and metadata stream names) are unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* style(windows): format signature admission regressions

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>

scripts/verify-windows-native-artifact.mjs

@@ -27,9 +27,9 @@ if (
   );
 const offset = bytes.length >= 64 ? bytes.readUInt32LE(0x3c) : bytes.length;
 if (
-  bytes.toString("ascii", 0, 2) !== "MZ" ||
+  bytes.toString("latin1", 0, 2) !== "MZ" ||
   offset > bytes.length - 26 ||
-  bytes.toString("ascii", offset, offset + 4) !== "PE\0\0" ||
+  bytes.toString("latin1", offset, offset + 4) !== "PE\0\0" ||
   bytes.readUInt16LE(offset + 4) !== 0x8664 ||
   bytes.readUInt16LE(offset + 24) !== 0x20b
 )

src/domain/native/nativeAotPe.test.ts

@@ -159,3 +159,22 @@ describe("read-only NativeAOT PE parser", () => {
     });
   });
 });
+
+describe("NativeAOT PE signature admission", () => {
+  it.each([
+    ["MZ", "Not a PE image"],
+    ["PE", "Invalid PE signature"],
+  ])(
+    "rejects a high-bit lookalike of the %s signature",
+    (signature, reason) => {
+      const bytes = nativeAotPeFixture();
+      const offset = signature === "MZ" ? 0 : bytes.readUInt32LE(0x3c);
+      for (const position of [offset, offset + 1])
+        bytes.writeUInt8(bytes.readUInt8(position) | 0x80, position);
+      expect(
+        parseNativeAotPe(bytes, nativeAotPeDigest(bytes), SUMMARY_BUDGET)
+          .summary,
+      ).toMatchObject({ status: "not_applicable", reason });
+    },
+  );
+});

src/domain/native/nativeAotPe.ts

@@ -384,11 +384,11 @@ class PeImage {
 
   constructor(bytes: Buffer) {
     this.#bytes = bytes;
-    if (bytes.length < 0x40 || bytes.toString("ascii", 0, 2) !== "MZ")
+    if (bytes.length < 0x40 || bytes.toString("latin1", 0, 2) !== "MZ")
       throw new Error("Not a PE image");
     const peOffset = this.#u32(0x3c);
     this.#range(peOffset, 24);
-    if (bytes.toString("ascii", peOffset, peOffset + 4) !== "PE\0\0")
+    if (bytes.toString("latin1", peOffset, peOffset + 4) !== "PE\0\0")
       throw new Error("Invalid PE signature");
     this.#machine = this.#u16(peOffset + 4);
     const count = this.#u16(peOffset + 6);

src/windows/WindowsNativeLoader.test.ts

@@ -143,6 +143,22 @@ describe("packaged Windows native admission", () => {
     },
   );
 
+  it.each([
+    { signature: "MZ", offset: 0 },
+    { signature: "PE", offset: 64 },
+  ])(
+    "rejects digest-valid high-bit $signature lookalikes before loading",
+    ({ offset }) => {
+      const value = fixture();
+      value.bytes[offset] = (value.bytes[offset] ?? 0) | 0x80;
+      value.manifest.artifactSha256 = createHash("sha256")
+        .update(value.bytes)
+        .digest("hex");
+      expect(loadWindowsNativeAuthority(value.host).available).toBe(false);
+      expect(value.loads()).toBe(0);
+    },
+  );
+
   it("rejects a digest-valid artifact for another machine", () => {
     const value = fixture();
     value.bytes.writeUInt16LE(0xaa64, 68);

src/windows/WindowsNativeLoader.ts

@@ -132,12 +132,12 @@ export const loadWindowsNativeAuthority = (
 };
 
 const assertWindowsX64Artifact = (bytes: Buffer): void => {
-  if (bytes.length < 64 || bytes.toString("ascii", 0, 2) !== "MZ")
+  if (bytes.length < 64 || bytes.toString("latin1", 0, 2) !== "MZ")
     throw new Error("Windows native artifact is not a PE image");
   const offset = bytes.readUInt32LE(0x3c);
   if (
     offset > bytes.length - 26 ||
-    bytes.toString("ascii", offset, offset + 4) !== "PE\0\0" ||
+    bytes.toString("latin1", offset, offset + 4) !== "PE\0\0" ||
     bytes.readUInt16LE(offset + 4) !== 0x8664 ||
     bytes.readUInt16LE(offset + 24) !== 0x20b
   )