Example of using hardware AES 256 Crypto in CBC mode on the ESP32 using ESP-IDF

esp32_aes_example.c

#include <string.h>
#include <stdio.h>
#include <hwcrypto/aes.h>

/* 
  For  Encryption time: 1802.40us  (9.09 MB/s) at 16kB blocks.
*/

static inline int32_t _getCycleCount(void) {
  int32_t ccount;
  __asm__ __volatile__("rsr %0,ccount":"=a" (ccount));
  return ccount;
}

char plaintext[16384];
char encrypted[16384];


int encodetest()
{
	uint8_t key[32];
	uint8_t iv[16];

  //If you have cryptographically random data in the start of your payload, you do not need
	//an IV.  If you start a plaintext payload, you will need an IV.
	memset( iv, 0, sizeof( iv ) );

	//Right now, I am using a key of all zeroes.  This should change.  You should fill the key
  //out with actual data. 
	memset( key, 0, sizeof( key ) );

	memset( plaintext, 0, sizeof( plaintext ) );
	strcpy( plaintext, "Hello, world, how are you doing today?" );

	//Just FYI - you must be encrypting/decrypting data that is in BLOCKSIZE chunks!!!

	esp_aes_context ctx;
	esp_aes_init( &ctx );
	esp_aes_setkey( &ctx, key, 256 );
	int32_t start = _getCycleCount();
	esp_aes_crypt_cbc( &ctx, ESP_AES_ENCRYPT, sizeof(plaintext), iv, (uint8_t*)plaintext, (uint8_t*)encrypted );
	int32_t end = _getCycleCount();
	float enctime = (end-start)/240.0;
	printf( "Encryption time: %.2fus  (%f MB/s)\n", enctime, (sizeof(plaintext)*1.0)/enctime );

	//See encrypted payload, and wipe out plaintext.
	memset( plaintext, 0, sizeof( plaintext ) );
	int i;
	for( i = 0; i < 128; i++ )
	{
		printf( "%02x[%c]%c", encrypted[i], (encrypted[i]>31)?encrypted[i]:' ', ((i&0xf)!=0xf)?' ':'\n' );
	}
	printf( "\n" );
	//Must reset IV.
	//XXX TODO: Research further: I found out if you don't reset the IV, the first block will fail
	//but subsequent blocks will pass.  Is there some strange cryptoalgebra going on that permits this?
	printf( "IV: %02x %02x\n", iv[0], iv[1] );
	memset( iv, 0, sizeof( iv ) );  


	//Use the ESP32 to decrypt the CBC block.
	esp_aes_crypt_cbc( &ctx, ESP_AES_DECRYPT, sizeof(encrypted), iv, (uint8_t*)encrypted, (uint8_t*)plaintext );


	//Verify output
	for( i = 0; i < 128; i++ )
	{
		printf( "%02x[%c]%c", plaintext[i],  (plaintext[i]>31)?plaintext[i]:' ', ((i&0xf)!=0xf)?' ':'\n' );
	}
	printf( "\n" );


	esp_aes_free( &ctx );
}

Hi,

how can you compile it please? can you give me the procedure to make it run on esp32?
thanks man

Hi,

how can you compile it please? can you give me the procedure to make it run on esp32?
thanks man

#include <string.h>
#include <stdio.h>
#include <hwcrypto/aes.h>

/*
For Encryption time: 1802.40us (9.09 MB/s) at 16kB blocks.
*/

static inline int32_t _getCycleCount(void) {
int32_t ccount;
asm volatile("rsr %0,ccount":"=a" (ccount));
return ccount;
}

char plaintext[16384];
char encrypted[16384];

int encodetest()
{
uint8_t key[32];
uint8_t iv[16];

//If you have cryptographically random data in the start of your payload, you do not need
//an IV. If you start a plaintext payload, you will need an IV.
memset( iv, 0, sizeof( iv ) );

//Right now, I am using a key of all zeroes. This should change. You should fill the key
//out with actual data.
memset( key, 0, sizeof( key ) );

memset( plaintext, 0, sizeof( plaintext ) );
strcpy( plaintext, "Hello, world, how are you doing today?" );

//Just FYI - you must be encrypting/decrypting data that is in BLOCKSIZE chunks!!!

esp_aes_context ctx;
esp_aes_init( &ctx );
esp_aes_setkey( &ctx, key, 256 );
int32_t start = _getCycleCount();
esp_aes_crypt_cbc( &ctx, ESP_AES_ENCRYPT, sizeof(plaintext), iv, (uint8_t*)plaintext, (uint8_t*)encrypted );
int32_t end = _getCycleCount();
float enctime = (end-start)/240.0;
Serial.printf( "Encryption time: %.2fus (%f MB/s)\n", enctime, (sizeof(plaintext)*1.0)/enctime );

//See encrypted payload, and wipe out plaintext.
memset( plaintext, 0, sizeof( plaintext ) );
int i;
for( i = 0; i < 128; i++ )
{
Serial.printf( "%02x[%c]%c", encrypted[i], (encrypted[i]>31)?encrypted[i]:' ', ((i&0xf)!=0xf)?' ':'\n' );
}
Serial.printf( "\n" );
//Must reset IV.
//XXX TODO: Research further: I found out if you don't reset the IV, the first block will fail
//but subsequent blocks will pass. Is there some strange cryptoalgebra going on that permits this?
Serial.printf( "IV: %02x %02x\n", iv[0], iv[1] );
memset( iv, 0, sizeof( iv ) );

//Use the ESP32 to decrypt the CBC block.
esp_aes_crypt_cbc( &ctx, ESP_AES_DECRYPT, sizeof(encrypted), iv, (uint8_t*)encrypted, (uint8_t*)plaintext );

//Verify output
for( i = 0; i < 128; i++ )
{
Serial.printf( "%02x[%c]%c", plaintext[i], (plaintext[i]>31)?plaintext[i]:' ', ((i&0xf)!=0xf)?' ':'\n' );
}
Serial.printf( "\n" );

esp_aes_free( &ctx );
}
void setup() {
// put your setup code here, to run once:
Serial.begin(115200);
encodetest();
}

void loop() {
// put your main code here, to run repeatedly:

}

@m0923678421 , can you give example how to generate base64 hash?

Note that the header file has moved and changed its name. It is now "aes/esp_aes.h"

One could also use "mbedtls/aes.h" though the names of the functions are a little different. This version of functions allows for a separate encrypt and decrypt key to be used in case you are mixing between these (so you don't have to reset the key each time) though the mbedtls code for AES is slightly larger code (~40 bytes more).