diff --git a/test/utility/frame.cpp b/test/utility/frame.cpp
index d79637aa8a..786ef0ba38 100644
--- a/test/utility/frame.cpp
+++ b/test/utility/frame.cpp
@@ -410,6 +410,43 @@ BOOST_AUTO_TEST_CASE( continuous_word_mask ) {
 	BOOST_CHECK_EQUAL( pkey_temp, frame::circshift_prepared_key(pkey,3) );
 }
 
+BOOST_AUTO_TEST_CASE( continuous_byte_mask ) {
+	uint8_t input[16];
+	uint8_t output[16];
+	
+	uint8_t masked[16] = {0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x00};
+
+	frame::masking_key_type key;
+	key.c[0] = 0x00;
+	key.c[1] = 0x01;
+	key.c[2] = 0x02;
+	key.c[3] = 0x03;
+
+	// One call
+	size_t pkey,pkey_temp;
+	pkey = frame::prepare_masking_key(key);
+	std::fill_n(input,16,0x00);
+	std::fill_n(output,16,0x00);
+	frame::byte_mask_circ(input,output,15,pkey);
+	BOOST_CHECK( std::equal(output,output+16,masked) );
+
+	// calls not split on word boundaries
+	pkey = frame::prepare_masking_key(key);
+	std::fill_n(input,16,0x00);
+	std::fill_n(output,16,0x00);
+	
+	pkey_temp = frame::byte_mask_circ(input,output,7,pkey);
+	BOOST_CHECK( std::equal(output,output+7,masked) );
+	BOOST_CHECK( pkey_temp == frame::circshift_prepared_key(pkey,3) );
+
+	pkey_temp = frame::byte_mask_circ(input+7,output+7,8,pkey_temp);
+	BOOST_CHECK( std::equal(output,output+16,masked) );
+	BOOST_CHECK_EQUAL( pkey_temp, frame::circshift_prepared_key(pkey,3) );
+}
+
 BOOST_AUTO_TEST_CASE( continuous_word_mask_inplace ) {
 	uint8_t buffer[16];
 	
@@ -444,6 +481,40 @@ BOOST_AUTO_TEST_CASE( continuous_word_mask_inplace ) {
 	BOOST_CHECK_EQUAL( pkey_temp, frame::circshift_prepared_key(pkey,3) );
 }
 
+BOOST_AUTO_TEST_CASE( continuous_byte_mask_inplace ) {
+	uint8_t buffer[16];
+	
+	uint8_t masked[16] = {0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x03,
+                          0x00, 0x01, 0x02, 0x00};
+
+	frame::masking_key_type key;
+	key.c[0] = 0x00;
+	key.c[1] = 0x01;
+	key.c[2] = 0x02;
+	key.c[3] = 0x03;
+
+	// One call
+	size_t pkey,pkey_temp;
+	pkey = frame::prepare_masking_key(key);
+	std::fill_n(buffer,16,0x00);
+	frame::byte_mask_circ(buffer,15,pkey);
+	BOOST_CHECK( std::equal(buffer,buffer+16,masked) );
+
+	// calls not split on word boundaries
+	pkey = frame::prepare_masking_key(key);
+	std::fill_n(buffer,16,0x00);
+	
+	pkey_temp = frame::byte_mask_circ(buffer,7,pkey);
+	BOOST_CHECK( std::equal(buffer,buffer+7,masked) );
+	BOOST_CHECK_EQUAL( pkey_temp, frame::circshift_prepared_key(pkey,3) );
+
+	pkey_temp = frame::byte_mask_circ(buffer+7,8,pkey_temp);
+	BOOST_CHECK( std::equal(buffer,buffer+16,masked) );
+	BOOST_CHECK_EQUAL( pkey_temp, frame::circshift_prepared_key(pkey,3) );
+}
+
 BOOST_AUTO_TEST_CASE( continuous_word_mask2 ) {	
 	uint8_t buffer[12] = {0xA6, 0x15, 0x97, 0xB9,
                           0x81, 0x50, 0xAC, 0xBA,
diff --git a/websocketpp/frame.hpp b/websocketpp/frame.hpp
index 202a1dddd3..0a09de5830 100644
--- a/websocketpp/frame.hpp
+++ b/websocketpp/frame.hpp
@@ -831,6 +831,58 @@ inline size_t word_mask_circ(uint8_t* data, size_t length, size_t prepared_key){
     return word_mask_circ(data,data,length,prepared_key);
 }
 
+/// Circular byte aligned mask/unmask
+/**
+ * Performs a circular mask/unmask in byte sized chunks using pre-prepared keys
+ * that store state between calls. Best for providing streaming masking or 
+ * unmasking of small chunks at a time of a larger message. Requires that the
+ * underlying allocated size of the data buffer be a multiple of the word size.
+ * Data in the buffer after `length` will be overwritten only with the same 
+ * values that were originally present.
+ * 
+ * word_mask returns a copy of prepared_key circularly shifted based on the 
+ * length value. The returned value may be fed back into byte_mask when more 
+ * data is available.
+ *
+ * @param data Character buffer to mask
+ *
+ * @param length Length of data
+ *
+ * @param prepared_key Prepared key to use.
+ *
+ * @return the prepared_key shifted to account for the input length
+ */
+inline size_t byte_mask_circ(uint8_t * input, uint8_t * output, size_t length, 
+    size_t prepared_key)
+{
+    uint32_converter key;
+    key.i = prepared_key;
+    
+    for (size_t i = 0; i < length; ++i) {
+        output[i] = input[i] ^ key.c[i % 4];
+    }
+    
+    return circshift_prepared_key(prepared_key,length % 4);
+}
+
+/// Circular byte aligned mask/unmask (in place)
+/**
+ * In place version of byte_mask_circ
+ *
+ * @see byte_mask_circ
+ *
+ * @param data Character buffer to read from and write to
+ *
+ * @param length Length of data
+ *
+ * @param prepared_key Prepared key to use.
+ *
+ * @return the prepared_key shifted to account for the input length
+ */
+inline size_t byte_mask_circ(uint8_t* data, size_t length, size_t prepared_key){
+    return byte_mask_circ(data,data,length,prepared_key);
+}
+
 } // namespace frame
 } // namespace websocketpp