Merge pull request #218 from pavestru/208-integrity-check-correlation-id

208 Integrity-check for the correlation ID

Author: mmacai

CHANGELOG.md

@@ -10,6 +10,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Added
 
 - [API Gateway] New custom metric: `rig_proxy_requests_total`. For details see [`metrics-details.md`](docs/metrics-details.md). [#157](https://github.com/Accenture/reactive-interaction-gateway/issues/157)
+- [Rig] Integrity-check for the correlation ID [#218](https://github.com/Accenture/reactive-interaction-gateway/pull/218)
 - _Beta_ - Added Apache Avro support for consumer and producer as well as Kafka Schema Registry.
 - [Docs] Added new set of topics in documentation about Api Gateway, even streams and scaling.
 - [Docs] Added examples section to documentation website.

apps/rig/config/config.exs

@@ -58,6 +58,10 @@ config :rig, Rig.EventStream.KafkaToHttp,
   targets: {:system, :list, "FIREHOSE_KAFKA_HTTP_TARGETS", []},
   group_id: {:system, "KAFKATOHTTP_KAFKA_GROUP_ID", "rig-kafka-to-http"}
 
+config :rig, Rig.Connection.Codec,
+  codec_secret_key: {:system, "NODE_COOKIE", nil},
+  codec_default_key: "7tsf4Y6GTOfPY1iDo4PqZA=="
+
 config :porcelain, driver: Porcelain.Driver.Basic
 
 import_config "#{Mix.env()}.exs"

apps/rig/lib/rig/connection/codec.ex

@@ -1,29 +1,31 @@
 defmodule Rig.Connection.Codec do
   @moduledoc """
   Encode and decode a connection token, e.g., for correlation.
-
-  TODO: sign the token - using binary_to_term without signature verification is a threat.
   """
+  use Rig.Config, [:codec_secret_key]
 
   @doc "Turn a pid into an url-encoded string."
   @spec serialize(pid) :: binary
   def serialize(pid) do
+    conf = config()
+    secret_key = conf.codec_secret_key || conf.codec_default_key
     pid
     |> :erlang.term_to_binary()
+    |> encrypt(secret_key)
     |> Base.url_encode64()
-
-    # TODO sign this
   end
 
   # ---
 
   @doc "Convert a serialized string back into a pid."
   @spec deserialize(binary) :: {:ok, pid} | {:error, :not_base64 | :invalid_term}
   def deserialize(base64_encoded) do
-    # TODO check signature here
-
+    conf = config()
+    secret_key = conf.codec_secret_key || conf.codec_default_key
     with {:ok, decoded_binary} <- decode64(base64_encoded) do
-      binary_to_term(decoded_binary)
+      decoded_binary
+      |> decrypt(secret_key)
+      |> binary_to_term()
     end
   end
 
@@ -38,6 +40,46 @@ defmodule Rig.Connection.Codec do
 
   # ---
 
+  @doc """
+  The function processes any string or number and returns exactly 16 byte binary.
+  The hash produced by this function can be used as key by both encrypt and decrypt functions.
+  """
+  @spec hash(binary) :: binary
+  def hash(key) do
+    :crypto.hash(:md5, :erlang.term_to_binary(key))
+  end
+
+  @doc """
+  Encrypts value using key and returns init. vector, ciphertag (MAC) and ciphertext concatenated.
+  Additional authenticated data (AAD) adds parameters like protocol version num. to MAC
+  """
+  @aad "AES256GCM"
+  @spec encrypt(binary, binary) :: binary
+  def encrypt(val, key) do
+    mode = :aes_gcm
+    secret_key = hash(key)
+    init_vector = :crypto.strong_rand_bytes(16)
+    {ciphertext, ciphertag} =
+      :crypto.block_encrypt(mode, secret_key, init_vector, {@aad, to_string(val), 16})
+    init_vector <> ciphertag <> ciphertext
+  end
+
+  # ---
+
+  @doc """
+  Decrypts ciphertext using key.
+  Ciphertext is init. vector, ciphertag (MAC) and the actual ciphertext concatenated.
+  """
+  @spec decrypt(binary, binary) :: binary
+  def decrypt(ciphertext, key) do
+    mode = :aes_gcm
+    secret_key = hash(key)
+    <<init_vector::binary-16, tag::binary-16, ciphertext::binary>> = ciphertext
+    :crypto.block_decrypt(mode, secret_key, init_vector, {@aad, ciphertext, tag})
+  end
+
+  # ---
+
   @spec decode64(binary()) :: {:ok, binary()} | {:error, :not_base64}
   defp decode64(base64) do
     case Base.url_decode64(base64) do

apps/rig/lib/rig/connection/codec_test.exs

@@ -5,7 +5,7 @@ defmodule Rig.ConnectionCodecTest do
 
   use ExUnitProperties
 
-  import Rig.Connection.Codec, only: [serialize: 1, deserialize!: 1]
+  import Rig.Connection.Codec, only: [serialize: 1, deserialize!: 1, encrypt: 2, decrypt: 2]
 
   property "Serialize and deserialize invert each other" do
     # `check all id <- integer(0..0x7FFF)` does not work on a single node.
@@ -29,4 +29,81 @@ defmodule Rig.ConnectionCodecTest do
       refute :c.pid(id, serial, creation) |> serialize() |> String.contains?(bad_chars)
     end
   end
+
+  property "Encryption / decryption works for key \"magiccookie\"" do
+    id = 0
+    key = "magiccookie"
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for key of empty string" do
+    id = 0
+    key = ""
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for key \"a\"" do
+    id = 0
+    key = "a"
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for a very long key (90 bytes)" do
+    id = 0
+    key = "QWErty1234QWErty1234QWErty1234QWErty1234QWErty1234QWErty1234QWErty1234QWErty1234"
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for key of integer 0" do
+    id = 0
+    key = 0
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for key of negative integer" do
+    id = 0
+    key = -123
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
+
+  property "Encryption / decryption works for key of large integer" do
+    id = 0
+    key = 12_345_678_901_234_567_890
+
+    check all serial <- integer(0..0x1FFF),
+              creation <- integer(0..0x03) do
+      pid = :c.pid(id, serial, creation) |> :erlang.term_to_binary
+      assert pid == pid |> encrypt(key) |> decrypt(key)
+    end
+  end
 end

docs/rig-ops-guide.md

@@ -58,7 +58,7 @@ Variable            
 `KINESIS_OTP_JAR` | Path to the `OtpErlang.jar` file that contains the `JInterface` implementation. If left empty, RIG picks the file from its Erlang environment (Erlang must be compiled with Java support enabled). | nil
 `KINESIS_STREAM` | The name of the Kinesis stream to consume. | "RIG-outbound"
 `LOG_LEVEL` | Controls logging level for RIG, available values are: "debug", "info", "warn", "error". Production is using "warn" level. | :warn
-`NODE_COOKIE` | Erlang cookie used in distributed mode, so nodes in cluster can communicate between each other. | nil
+`NODE_COOKIE` | Erlang cookie used in distributed mode, so nodes in cluster can communicate between each other.<br />Used also as secret key for integrity-check of correlation IDs. | nil
 `NODE_HOST` | Erlang hostname for given node, used to build Erlang long-name `rig@NODE_HOST`. This value is used by Erlang's distributed mode, so nodes can see each other. | nil
 `PROXY_CONFIG_FILE` | Configuration JSON file with initial API definition for API Proxy. Use this variable to pass either a path to a JSON file, or the JSON string itself. A path can be given in absolute or in relative form (e.g., `proxy/your_json_file.json`). If given in relative form, the working directory is one of RIG's `priv` dirs (e.g., `/opt/sites/rig/lib/rig_inbound_gateway-2.0.2/priv/` in a Docker container). | nil
 `PROXY_HTTP_ASYNC_RESPONSE_TIMEOUT` | In case an endpoint has `target` set to `http` and `response_from` set to `http_async`, this is the maximum delay between an HTTP request and the corresponding async HTTP response message. | 5000