diff --git a/configs/config-psa-crypto.h b/configs/config-psa-crypto.h
index 9c4f62a..9ae09c9 100644
--- a/configs/config-psa-crypto.h
+++ b/configs/config-psa-crypto.h
@@ -398,13 +398,46 @@
 /**
  * \def MBEDTLS_AES_ROM_TABLES
  *
- * Store the AES tables in ROM.
+ * Use precomputed AES tables stored in ROM.
  *
- * Uncomment this macro to store the AES tables in ROM.
+ * Uncomment this macro to use precomputed AES tables stored in ROM.
+ * Comment this macro to generate AES tables in RAM at runtime.
+ *
+ * Tradeoff: Using precomputed ROM tables reduces RAM usage by ~8kb
+ * (or ~2kb if \c MBEDTLS_AES_FEWER_TABLES is used) and reduces the
+ * initialization time before the first AES operation can be performed.
+ * It comes at the cost of additional ~8kb ROM use (resp. ~2kb if \c
+ * MBEDTLS_AES_FEWER_TABLES below is used), and potentially degraded
+ * performance if ROM access is slower than RAM access.
+ *
+ * This option is independent of \c MBEDTLS_AES_FEWER_TABLES.
+ *
  */
 //#define MBEDTLS_AES_ROM_TABLES
 
 /**
+ * \def MBEDTLS_AES_FEWER_TABLES
+ *
+ * Use less ROM/RAM for AES tables.
+ *
+ * Uncommenting this macro omits 75% of the AES tables from
+ * ROM / RAM (depending on the value of \c MBEDTLS_AES_ROM_TABLES)
+ * by computing their values on the fly during operations
+ * (the tables are entry-wise rotations of one another).
+ *
+ * Tradeoff: Uncommenting this reduces the RAM / ROM footprint
+ * by ~6kb but at the cost of more arithmetic operations during
+ * runtime. Specifically, one has to compare 4 accesses within
+ * different tables to 4 accesses with additional arithmetic
+ * operations within the same table. The performance gain/loss
+ * depends on the system and memory details.
+ *
+ * This option is independent of \c MBEDTLS_AES_ROM_TABLES.
+ *
+ */
+//#define MBEDTLS_AES_FEWER_TABLES
+
+/**
  * \def MBEDTLS_CAMELLIA_SMALL_MEMORY
  *
  * Use less ROM for the Camellia implementation (saves about 768 bytes).
@@ -470,6 +503,7 @@
 #define MBEDTLS_ECP_DP_BP384R1_ENABLED
 #define MBEDTLS_ECP_DP_BP512R1_ENABLED
 #define MBEDTLS_ECP_DP_CURVE25519_ENABLED
+#define MBEDTLS_ECP_DP_CURVE448_ENABLED
 
 /**
  * \def MBEDTLS_ECP_NIST_OPTIM
@@ -1704,6 +1738,26 @@
 //#define MBEDTLS_PLATFORM_NV_SEED_READ_MACRO   mbedtls_platform_std_nv_seed_read /**< Default nv_seed_read function to use, can be undefined */
 //#define MBEDTLS_PLATFORM_NV_SEED_WRITE_MACRO  mbedtls_platform_std_nv_seed_write /**< Default nv_seed_write function to use, can be undefined */
 
+/**
+ * Uncomment the macro to let mbed TLS use your alternate implementation of
+ * mbedtls_platform_zeroize(). This replaces the default implementation in
+ * platform_util.c.
+ *
+ * mbedtls_platform_zeroize() is a widely used function across the library to
+ * zero a block of memory. The implementation is expected to be secure in the
+ * sense that it has been written to prevent the compiler from removing calls
+ * to mbedtls_platform_zeroize() as part of redundant code elimination
+ * optimizations. However, it is difficult to guarantee that calls to
+ * mbedtls_platform_zeroize() will not be optimized by the compiler as older
+ * versions of the C language standards do not provide a secure implementation
+ * of memset(). Therefore, MBEDTLS_PLATFORM_ZEROIZE_ALT enables users to
+ * configure their own implementation of mbedtls_platform_zeroize(), for
+ * example by using directives specific to their compiler, features from newer
+ * C standards (e.g using memset_s() in C11) or calling a secure memset() from
+ * their system (e.g explicit_bzero() in BSD).
+ */
+//#define MBEDTLS_PLATFORM_ZEROIZE_ALT
+
 /* \} name SECTION: Customisation configuration options */
 
 #include "mbedtls/check_config.h"