diff --git a/src/ws/dune b/src/ws/dune
index 7fc4745c..2ac87ea4 100644
--- a/src/ws/dune
+++ b/src/ws/dune
@@ -4,4 +4,8 @@
   (public_name tiny_httpd_ws)
   (synopsis "Websockets for tiny_httpd")
   (private_modules common_ utils_)
+  (foreign_stubs
+    (language c)
+    (names tiny_httpd_ws_stubs)
+    (flags :standard -std=c99 -fPIC -O2))
   (libraries tiny_httpd threads))
diff --git a/src/ws/tiny_httpd_ws.ml b/src/ws/tiny_httpd_ws.ml
index fb8e7af6..c0cd7b2e 100644
--- a/src/ws/tiny_httpd_ws.ml
+++ b/src/ws/tiny_httpd_ws.ml
@@ -272,14 +272,14 @@ module Reader = struct
           self.header.payload_len self.header.mask);*)
     ()
 
+  external apply_masking_ : bytes -> bytes -> int -> int -> unit
+    = "tiny_httpd_ws_apply_masking"
+    [@@noalloc]
   (** Apply masking to the parsed data *)
-  let apply_masking ~mask_key (buf : bytes) off len : unit =
-    for i = 0 to len - 1 do
-      let c = Bytes.get buf (off + i) in
-      let c_m = Bytes.unsafe_get mask_key (i land 0b11) in
-      let c_xor = Char.chr (Char.code c lxor Char.code c_m) in
-      Bytes.set buf (off + i) c_xor
-    done
+
+  let[@inline] apply_masking ~mask_key (buf : bytes) off len : unit =
+    assert (off >= 0 && off + len <= Bytes.length buf);
+    apply_masking_ mask_key buf off len
 
   let read_body_to_string (self : t) : string =
     let len = self.header.payload_len in
diff --git a/src/ws/tiny_httpd_ws_stubs.c b/src/ws/tiny_httpd_ws_stubs.c
new file mode 100644
index 00000000..779e255a
--- /dev/null
+++ b/src/ws/tiny_httpd_ws_stubs.c
@@ -0,0 +1,21 @@
+
+#include <caml/alloc.h>
+#include <caml/memory.h>
+#include <caml/mlvalues.h>
+
+CAMLprim value tiny_httpd_ws_apply_masking(value _mask_key, value _buf,
+                                           value _offset, value _len) {
+  CAMLparam4(_mask_key, _buf, _offset, _len);
+
+  char const *mask_key = String_val(_mask_key);
+  char *buf = Bytes_val(_buf);
+  intnat offset = Int_val(_offset);
+  intnat len = Int_val(_len);
+
+  for (intnat i = 0; i < len; ++i) {
+    char c = buf[offset + i];
+    char c_m = mask_key[i & 0x3];
+    buf[offset + i] = c ^ c_m;
+  }
+  CAMLreturn(Val_unit);
+}