diff --git a/src/backend_libsecp256k1/libsecp256k1.nim b/src/backend_libsecp256k1/libsecp256k1.nim
index 2f72407..84320e6 100644
--- a/src/backend_libsecp256k1/libsecp256k1.nim
+++ b/src/backend_libsecp256k1/libsecp256k1.nim
@@ -20,13 +20,13 @@ proc `=destroy`(ctx: ptr secp256k1_context) =
     ctx.secp256k1_context_destroy
 
 type
-  Serialized_PubKey = ByteArrayBE[65]
+  Serialized_PubKey = array[65, byte]
 
 proc asPtrPubKey(key: PublicKey): ptr secp256k1_pubkey =
-  cast[ptr secp256k1_pubkey](unsafeAddr key.raw_key)
+  cast[ptr secp256k1_pubkey](unsafeAddr key)
 
 proc asPtrCuchar(key: PrivateKey): ptr cuchar =
-  cast[ptr cuchar](unsafeAddr key.raw_key)
+  cast[ptr cuchar](unsafeAddr key)
 
 proc asPtrCuchar(key: Serialized_PubKey): ptr cuchar =
   cast[ptr cuchar](unsafeAddr key)
diff --git a/src/datatypes.nim b/src/datatypes.nim
index 1c69cbc..f885a21 100644
--- a/src/datatypes.nim
+++ b/src/datatypes.nim
@@ -7,22 +7,43 @@
 #
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
-import ./private/lowlevel_types
-import ttmath
+import ./private/conversion_bytes
+export toHex, hexToByteArrayBE, hexToSeqByteBE
 
-export lowlevel_types, ttmath
 
+# Note: Fields are intentionally kept private
 type
   PublicKey* = object
-    raw_key*: ByteArrayBE[64]
+    Fraw_key: array[64, byte]
 
   PrivateKey* = object
-    raw_key*: ByteArrayBE[32]
-    public_key*: PublicKey
-
-  BaseKey* = PrivateKey|PublicKey
+    Fraw_key: array[32, byte]
+    Fpublic_key: PublicKey
 
   Signature* {.packed.}= object
-    r*: UInt256
-    s*: UInt256
-    v*: range[0.byte .. 1.byte]
+    Fr: array[32, byte]
+    Fs: array[32, byte]
+    Fv: range[0.byte .. 1.byte]
+
+
+# "Public" accessors, only exposed to internal modules
+
+template genAccessors(name: untyped, fieldType, objType: typedesc): untyped =
+  # Access
+  proc name*(obj: objType): fieldType {.noSideEffect, inline, noInit.} =
+    obj.`F name`
+
+  # Assignement
+  proc `name=`*(obj: var objType, value: fieldType): fieldType {.noSideEffect, inline.} =
+    obj.`F name` = value
+
+  # Mutable
+  proc `name`*(obj: var objType): var fieldType {.noSideEffect, inline.} =
+    obj.`F name`
+
+genAccessors(raw_key, array[64, byte], PublicKey)
+genAccessors(raw_key, array[32, byte], PrivateKey)
+genAccessors(public_key, PublicKey, PrivateKey)
+genAccessors(s, array[32, byte], Signature)
+genAccessors(r, array[32, byte], Signature)
+genAccessors(v, range[0.byte .. 1.byte], Signature)
diff --git a/src/eth_keys.nim b/src/eth_keys.nim
index 19b57f5..417d787 100644
--- a/src/eth_keys.nim
+++ b/src/eth_keys.nim
@@ -7,13 +7,12 @@
 #
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
-import  ./datatypes,
-        ./datatypes_interface
+import ./datatypes
+export PublicKey, PrivateKey, Signature
 
+import ./datatypes_interface
+export datatypes_interface
 
-export  datatypes,
-        datatypes_interface
-
-
-import ttmath
-export ttmath
+when defined(backend_native):
+  import ttmath
+  export ttmath
diff --git a/src/private/lowlevel_types.nim b/src/private/conversion_bytes.nim
similarity index 70%
rename from src/private/lowlevel_types.nim
rename to src/private/conversion_bytes.nim
index 2ab5d60..941a117 100644
--- a/src/private/lowlevel_types.nim
+++ b/src/private/conversion_bytes.nim
@@ -7,7 +7,7 @@
 #
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
-import ttmath, strutils
+import strutils
 
 # Note on endianness:
 # - UInt256 uses host endianness
@@ -19,24 +19,7 @@ import ttmath, strutils
 #   https://www.reddit.com/r/crypto/comments/6287my/explanations_on_the_keccaksha3_paddingbyte/
 #   Note: Since Nim's Keccak-Tiny only accepts string as input, endianness does not matter.
 
-type ByteArrayBE*[N: static[int]] = array[N, byte]
-  ## A byte array that stores bytes in big-endian order
-
-proc readUint256BE*(ba: ByteArrayBE[32]): UInt256 {.noSideEffect, inline.}=
-  ## Convert a big-endian array of Bytes to an UInt256 (in native host endianness)
-  const N = 32
-  for i in 0 ..< N:
-    {.unroll: 4.}
-    result = result shl 8 or ba[i].u256
-
-proc toByteArrayBE*(num: UInt256): ByteArrayBE[32] {.noSideEffect, noInit, inline.}=
-  ## Convert an UInt256 (in native host endianness) to a big-endian byte array
-  const N = 32
-  for i in 0 ..< N:
-    {.unroll: 4.}
-    result[i] = byte getUInt(num shr uint((N-1-i) * 8))
-
-proc readHexChar(c: char): byte {.noSideEffect.}=
+proc readHexChar*(c: char): byte {.noSideEffect.}=
   ## Converts an hex char to a byte
   case c
   of '0'..'9': result = byte(ord(c) - ord('0'))
@@ -45,7 +28,7 @@ proc readHexChar(c: char): byte {.noSideEffect.}=
   else:
     raise newException(ValueError, $c & "is not a hexademical character")
 
-proc skip0xPrefix(hexStr: string): int {.inline.} =
+proc skip0xPrefix*(hexStr: string): int {.inline.} =
   ## Returns the index of the first meaningful char in `hexStr` by skipping
   ## "0x" prefix
   if hexStr[0] == '0' and hexStr[1] in {'x', 'X'}:
@@ -68,7 +51,7 @@ proc hexToByteArrayBE*(hexStr: string, output: var openArray[byte]) {.inline.} =
   ## Read a hex string and store it in a Byte Array `output` in Big-Endian order
   hexToByteArrayBE(hexStr, output, 0, output.high)
 
-proc hexToByteArrayBE*[N: static[int]](hexStr: string): ByteArrayBE[N] {.noSideEffect, noInit, inline.}=
+proc hexToByteArrayBE*[N: static[int]](hexStr: string): array[N, byte] {.noSideEffect, noInit, inline.}=
   ## Read an hex string and store it in a Byte Array in Big-Endian order
   hexToByteArrayBE(hexStr, result)
 
@@ -83,33 +66,6 @@ proc hexToSeqByteBE*(hexStr: string): seq[byte] {.noSideEffect.}=
     result[i] = hexStr[2*i].readHexChar shl 4 or hexStr[2*i+1].readHexChar
     inc(i)
 
-proc hexToUInt256*(hexStr: string): UInt256 {.noSideEffect.}=
-  ## Read an hex string and store it in a UInt256
-  const N = 32
-
-  var i = skip0xPrefix(hexStr)
-
-  assert hexStr.len - i == 2*N
-
-  while i < 2*N:
-    result = result shl 4 or hexStr[i].readHexChar.uint.u256
-    inc(i)
-
-proc toHex*(n: UInt256): string {.noSideEffect.}=
-  ## Convert uint256 to its hex representation
-  ## Output is in lowercase
-
-  var rem = n # reminder to encode
-
-  const
-    N = 32 # nb of bytes in n
-    hexChars = "0123456789abcdef"
-
-  result = newString(2*N)
-  for i in countdown(2*N - 1, 0):
-    result[i] = hexChars[(rem and 0xF.u256).getUInt.int]
-    rem = rem shr 4
-
 proc toHexAux(ba: openarray[byte]): string {.noSideEffect.} =
   ## Convert a byte-array to its hex representation
   ## Output is in lowercase
@@ -132,7 +88,7 @@ proc toHex*(ba: openarray[byte]): string {.noSideEffect, inline.} =
   ##     - It is resistant against timing attack
   toHexAux(ba)
 
-proc toHex*(ba: ByteArrayBE): string {.noSideEffect, inline.} =
+proc toHex*[N: static[int]](ba: array[N, byte]): string {.noSideEffect, inline.} =
   ## Convert a byte-array to its hex representation
   ## Output is in lowercase
   ##
diff --git a/src/private/conversion_ttmath.nim b/src/private/conversion_ttmath.nim
new file mode 100644
index 0000000..2090f92
--- /dev/null
+++ b/src/private/conversion_ttmath.nim
@@ -0,0 +1,63 @@
+# Nim Eth-keys
+# Copyright (c) 2018 Status Research & Development GmbH
+# Licensed under either of
+#
+#  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0)
+#  * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT)
+#
+# at your option. This file may not be copied, modified, or distributed except according to those terms.
+
+import  ttmath, strutils,
+        conversion_bytes
+
+
+# Note on endianness:
+# - UInt256 uses host endianness
+# - Libsecp256k1, Ethereum EVM expect Big Endian
+#   https://github.com/ethereum/evmjit/issues/91
+# - Keccak expects least-significant byte first: http://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf
+#   Appendix B.1 p37 and outputs a hash with the same endianness as input
+#   http://www.dianacoman.com/2018/02/08/eucrypt-chapter-9-byte-order-and-bit-disorder-in-keccak/
+#   https://www.reddit.com/r/crypto/comments/6287my/explanations_on_the_keccaksha3_paddingbyte/
+#   Note: Since Nim's Keccak-Tiny only accepts string as input, endianness does not matter.
+
+proc toByteArrayBE*(num: UInt256): array[32, byte] {.noSideEffect, noInit, inline.}=
+  ## Convert an UInt256 (in native host endianness) to a big-endian byte array
+  const N = 32
+  for i in 0 ..< N:
+    {.unroll: 4.}
+    result[i] = byte getUInt(num shr uint((N-1-i) * 8))
+
+proc readUint256BE*(ba: array[32, byte]): UInt256 {.noSideEffect, inline.}=
+  ## Convert a big-endian array of Bytes to an UInt256 (in native host endianness)
+  const N = 32
+  for i in 0 ..< N:
+    {.unroll: 4.}
+    result = result shl 8 or ba[i].u256
+
+proc hexToUInt256*(hexStr: string): UInt256 {.noSideEffect.}=
+  ## Read an hex string and store it in a UInt256
+  const N = 32
+
+  var i = skip0xPrefix(hexStr)
+
+  assert hexStr.len - i == 2*N
+
+  while i < 2*N:
+    result = result shl 4 or hexStr[i].readHexChar.uint.u256
+    inc(i)
+
+proc toHex*(n: UInt256): string {.noSideEffect.}=
+  ## Convert uint256 to its hex representation
+  ## Output is in lowercase
+
+  var rem = n # reminder to encode
+
+  const
+    N = 32 # nb of bytes in n
+    hexChars = "0123456789abcdef"
+
+  result = newString(2*N)
+  for i in countdown(2*N - 1, 0):
+    result[i] = hexChars[(rem and 0xF.u256).getUInt.int]
+    rem = rem shr 4
diff --git a/tests/test_hex_bytes_conversion.nim b/tests/test_hex_bytes_conversion.nim
index e33a691..ead4234 100644
--- a/tests/test_hex_bytes_conversion.nim
+++ b/tests/test_hex_bytes_conversion.nim
@@ -7,8 +7,8 @@
 #
 # at your option. This file may not be copied, modified, or distributed except according to those terms.
 
-import  ../src/private/lowlevel_types
-import unittest, ttmath, strutils
+import  ../src/private/[conversion_bytes, conversion_ttmath]
+import unittest, ttmath, strutils # TODO remove ttmath needs if backend libsecp256k1
 
 
 suite "Testing conversion functions: Hex, Bytes, Endianness":