FNR Cipher

FNR Cipher is a Java implementation for Flexible Naor and Reingold encryption scheme.

---

It's simple. If you give an integer, you get a encrypted integer. This is two-way operation. That means, if you give the encrypted integer also you get the original integer. The FNR algorithm preserves your data size, no expand, no shrink. All operation space limited by key bit length.

This method also known as format preserving encryption. FNR algorithm is useful for small data types (up to 128 bits) such as credit card or user ids.

FNR uses AES-128 internally in each encryption/decryption rounds. FNR Java contains optimized and minimal AES-128 pure Java cipher. It's may be prefer instead of Java Cryptography Extension in case of performance considerations or JVM limitations. This library doesn't have any dependencies and it is fit for limited envorinments like Andorid. (The JAR file only 43kb.)

The FNR Java library is binary compatible with reference C implementation.

FNR Java provides some built-in codecs for basic data types encryption.

---

Built-In Supported Java Types:

Java Type	Codec	Notes
Byte	FNRCodec.BYTE	NP_SIGN codec not preserve sign.
Short	FNRCodec.SHORT	NP_SIGN codec not preserve sign.
Character	FNRCodec.CHAR	NP_SIGN codec not preserve sign.
Integer	FNRCodec.INT	NP_SIGN codec not preserve sign.
Float	FNRCodec.FLOAT	NP_SIGN_EXP codec not preserve sign and exponent.
Long	FNRCodec.LONG	NP_SIGN codec not preserve sign.
Double	FNRCodec.DOUBLE	NP_SIGN_EXP codec not preserve sign and exponent.
BigInteger	FNRCodec.BIGINT_128	The acceptable value range are -2^127 to 2^127-1 or 0 to 2^128-1
Date	FNRCodec.DATE	-
Inet4Address	FNRCodec.IPV4	-
Inet6Address	FNRCodec.IPV6

Note: All numeric codecs run as litte-endian for compatibiliy with other platform like C or Go and preserve sign and exponents as default.

Install

<dependency>
<groupId>io.octa.security</groupId>
<artifactId>fnr</artifactId>
<version>1.0.1</version>
</dependency>

Usage

Stringpassphrase = "this is a password";
Stringsalt = "this is a salt value"; // for built-in PBKDF2 key generation.byte[] aesKey = FNRUtils.createAes128KeyWithPBKDF2(passphrase, salt);
// Integer encryptionFNRKeykey = newFNRKey(aesKey, FNRCodec.INT.getRequiredKeyNumBits());
FNRTweaktweak = key.generateTweak("this is a tweak value");
intraw = 42;
intencrypted = FNRCipher.encrypt(FNRCodec.INT, key, tweak, raw);
intdecrypted = FNRCipher.decrypt(FNRCodec.INT, key, tweak, encrypted);
System.out.println("raw: " + raw); // prints 42System.out.println("encrypted: " + encrypted); // prints 1432569698System.out.println("decrypted: " + decrypted); // prints 42// IP encryptionkey = newFNRKey(aesKey, FNRCodec.IPV4.getRequiredKeyNumBits());
tweak = key.generateTweak("this is a tweak value");
Inet4AddressrawIP = (Inet4Address) Inet4Address.getByName("8.4.4.2");
Inet4AddressencryptedIP = FNRCipher.encrypt(FNRCodec.IPV4, key, tweak, rawIP);
Inet4AddressdecryptedIP = FNRCipher.decrypt(FNRCodec.IPV4, key, tweak, encryptedIP);
System.out.println("raw: " + rawIP); // prints 8.4.4.2System.out.println("encrypted: " + encryptedIP); // prints 25.123.159.248System.out.println("decrypted: " + decryptedIP); // prints 8.4.4.2

Performance

Library	AES Encryption Method	Encryption	Decryption	Notes
Reference C implementaion	OpenSSL	229141.720 ops/s	230386.135 ops/s	OpenSSL uses CPU AES Extension
FNR Java	Built-In	198160.740 ops/s	202775.251 ops/s	AES encryption with built-in minimal, optimized cipher
Java binding for Reference C implementaion	OpenSSL	105766.458 ops/s	106495.132 ops/s	I think the cause of bottleneck is JNI round-trip overhead.
FNR Java	JCE	82998.094 ops/s	81175.897 ops/s	AES encryption with standard Java Cryptography Extension

Tested on Intel(R) Core(TM) i7-4700MQ CPU @ 2.40GHz.

Java Benchmark: FNRCipherBenchmarkTest.java

JMH 1.11.3 (released 3 days ago)
VM version: JDK 1.8.0_66, VM 25.66-b17
VM invoker: /usr/lib/jvm/java-8-oracle/jre/bin/java
VM options: <none>
Warmup: 5 iterations, 1 s each
Measurement: 5 iterations, 1 s each
Timeout: 10 min per iteration
Threads: 1 thread, will synchronize iterations
Benchmark mode: Throughput, ops/time

C Benchmark: bench.c

GCC v5.2.1 compiles with -O2 and use OpenSSL 1.0.2d

FNR is designed by Sashank Dara (sadara@cisco.com), Scott Fluhrer (sfluhrer@cisco.com).

Java implementation was written by Mehmet Gurevin (mehmet.gurevin@octabase.com)

Copyright (c) 2015-2016, Octabase, Ltd. All Rights Reserved.