IEEE Consumer Electronics Magazine - May 2018 - 104

Hardware Matters

Obfuscated JPEG Image
Decompression IP Core
for Protecting Against
Reverse Engineering
By Dipanjan Roy and Anirban Sengupta

oint Photographic Experts Group
(JPEG) compression and decompression (CODEC) plays a crucial
role for all digital camera devices
in consumer electronics (CE) products.
Conventional JPEG CODEC design
techniques are nontrivial and vulnerable
to external threats of reverse engineering that result in cloning, counterfeiting,
and piracy. This article presents a novel
design of an obfuscated JPEG decompression intellectual property (IP) core
that enhances the complexity of reverse
engineering attacks from the perspective of a rogue element present anywhere in the integrated circuit design
flow. The presented IP core design introduces a simplified model for calculating
the digital pixel intensity values of a
decompressed image during image
reconstruction as well as protecting the
design against IP cloning. The proposed
obfuscated JPEG decompression IP
core is capable of achieving the desired
results in terms of strength of obfuscation, compression efficiency, mean-square
error (MSE), and peak signal-to-noise
ratio (PSNR) for various images obtained
from NASA and the Cornell University
public database.

OBFUSCATION OVERVIEW
JPEG image CODEC through hardware/
software has been commonly used in
digital images and videos for such CE
Digital Object Identifier 10.1109/MCE.2018.2797620
Date of publication: 10 April 2018

104 IEEE Consumer Electronics Magazine

may 2018

products as tablets, smartphones, and
digicams. Due to the advancement of
these imaging systems, which include
lens and recording and display components, the optical resolution of an image
has also increased. Higher optical resolution represents more detailed information
about an image, which results in a larger
image size. Higher image size results in
a critical problem for storage space and
transmission bandwidth.
The JPEG compression software
technique is one of the most commercially successful approaches for relieving
this crisis by reducing the image size.
However, the software design process of
JPEG CODEC is nontrivial and requires
significant effort for obtaining compressed images. Even the standard hardware-based approach is complex and not
resilient against externals threats, such as
reverse engineering attacks.
For CE devices, the JPEG CODEC
process can be efficiently performed by a
dedicated reusable IP core. However, due
to threats like IP cloning, IP infringement,
malicious Trojan insertion, and reverse
engineering, among others, IP-core-based
JPEG CODEC [6] alone is not sufficient,
as it necessitates protection [1]. IP watermarking, IP fingerprinting, computational
forensic engineering, and IP metering are
some of the passive protection techniques
that are only capable of tracing the illegal
copies of an IP core but not preventing it
from being counterfeit.
Obfuscation is an active IP core protection technique that hides the func-

tionality and implementation of an IP
core while passing through the different, but potentially untrustworthy, phases of the design flow. Obfuscation is a
process of transforming an original
application or design into its functionally equivalent form to make the reverse
engineering process significantly more
challenging [2]. All of the approaches
present in the literature perform JPEG
CODEC but do not design an obfuscated IP core for enhancing complexity
against attacks. In this article, we propose a hardware obfuscated IP core for
the JPEG decompression process that,
through high-level transformation, provides protection against reverse engineering attacks.

BACKGROUND ON SOFTWARE
JPEG IMAGE CODEC
The JPEG process that centers on the
discrete cosine transformation (DCT)
is one of the most popularly used lossy
image compression techniques in software. Two-dimensional (2-D) DCT
(2-D-DCT)-based JPEG operates on an
8 # 8 image block. Therefore, the input
image is divided into multiple nonoverlapping 8 # 8 blocks, and then
each block is transformed in parallel.
As DCT can be applied on pixel values
ranging from −128 to +127, each block
is leveled off by subtracting 128 from
each pixel value before performing
transformation. In JPEG compression
used in software, an N # N matrix is
obtained from a 2-D, square, grayscale

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - May 2018