

Proceedings of Engineering and Technology Innovation , vol. 3, 2016, pp. 19 - 21

Flexible Macroblock Ordering Scramble Encryption

Techniques for H.264/AVC Videos

Yih-Chuan Lin
1,*

, Wei-Siang Wang
1
, Yao-Tang Chang

1
Department of Computer Science and Information Engineering, National Formosa University,

Yunlin, Taiwan.

2
Department of Information Technology , Kao Yuan University, Kaohsiung, Taiwan.

Received 01 February 2016; received in revised form 12 March 2016; accept ed 02 April 2016

Abstract

In this paper, a ne w v ideo encryption method

through scramb ling the co mpressed videos is

presented, which is targeted for H.264/A VC

video encryption in order to provide the greatest

content protection in the compressed domain.

The proposed algorithm uses a macroblock

switching mechanism to scra mble the video

content for the compressed video sequences .

Only the instantaneous decoded reference (IDR)

pictures are scrambled to take the advantage of

drift e rror propagation fro m a ll the in-

ter-prediction fra mes. The proposed encryption

technique is designed to perform after the en-

coding process on the compressed bitstream

such that bitstream is modified by the parser

directly. Fina lly, the scra mbled video is still

format-co mpliant to a general H.264/A VC d e-

coder and can only be recovered by the author-

ized user that owns private key. Based on ex-

perimental results, the scene in the scrambled

video can be effective ly protected by the pro-

posed scheme with low co mputational co m-

plexity and negligible bitrate overhead.

Ke ywor ds : vio lation of privacy, H.264/A VC,

video encryption, protection effect

1. Introduction

With the rapid advance of mu ltimed ia

communicat ion and network technology, digita l

video applications are easy to be found every-

where. The comp ressed bitstream is transmitted

over the internet, decoded on a c lient and

showed. Digital v ideo surveillance and applica-

tions can be found ubiquitously in our daily lives.

For manage ment purposes, the captured videos

are usually encoded and transmitted over the

Internet to a third-party video service provider.

Vio lation of personal privacy and possible

leakage o f v ideo scene are addressed in the study.

In the case of video surveillance system, in order

to archive a huge surveillance data that gener-

ated fro m each IP ca me ra, the captured video

usually trans mitted to third-party servers . These

servers are so-called delegates that provide high

capacity for storage; reliable bandwidth re-

sources for end users to access; and simp le

manage ment interface fo r video owners. Ho w-

ever, there may have some untrustworthy system

administrators that take a pee k at the uploaded

video content in third-party service provider.

Selective encryption is a particu lar tech-
nique to encrypt the video content; only sens i-

tive part of the bitstream is encrypted. For ex-

amp le, the stream c ipher is just applied to ce rtain

parts that are sensitive for the video scene; s e-

lected bits are used XOR bit wise operation with

cipherte xt that generated by stream c ipher.

Shahid et al. [1] proposed a format-co mp liant

selective encryption method. For CA VLC, signs

of trailing ones and magnitude’s level suffix in

the specified range are encrypted. For CABAC,

the magnitude’s EG0 sub-suffix in the specified

range and signs of non-zero quantized transform

residual are encrypted. It is no change in final

bitrate. Wang et al. [2] proposed a tunable en-

cryption scheme; besides the above-mentioned,

signs of motion vector and intra prediction mode

are also encrypted. Obviously, the correspond-

ing computational cost is relatively lowe r than

full encryption.

2. Method

In this study, we imple ment two video en-

cryption algorith ms that prevent the sensitive

* Corresponding author. Email: lyc@nfu.edu.t w

Proceedings of Engineering and Technology Innovation , vol. 3, 2016, pp. 19 - 21

content fro m un-authorized access in the un-

trustworthy cloud. The encrypted video only

recovered from the trusted person who owns a

privacy key. The proposed s ystem arch itecture is

shown in Fig. 1. There e xist two video scram-

bling mechanisms that can be used to satisfy

diffe rent situations in the video encryption pro-

cess .

Fig. 1 Proposed system scenario

The proposed video scrambling methods are

separately described as follows :

2.1. Scrambling with other Videos

In a video surveillance environ ment, the

surveillance ca me ra is wide ly d istributed over

the building. The first scheme is to scramb le

mu lti-way v ideo contents simu ltaneously ; every

sequence contains part of pictures that belong to

the others . The coded IDR pictures in each s e-

quence are denoted by V = {v1, v2, ..., vi}, where i

is the number of input sequences. The ith coded

IDR p icture can be represented by vi = {vi1,

vi2, ..., vij}, where j is the number of slice groups.

After the encryption process, the scrambled IDR

picture is turned into }ˆ,...,ˆ,ˆ{ˆ kjk2k1i vvvv  , where 1

≤ k ≤ i.

(a) Generate the H.264/AVC bitstream: Each

came ra contains an H.264/A VC encoder, the

FMO configurations are listed below:

 The FMO map type is dispersed mode.

 4 slice groups in a coded picture.

(b) Detect the IDR picture: The detector parses
the input bitstream and looks for the NALU

type to indicate whether the IDR p icture has

happened or not. If the IDR picture is de-

tected, the sentinel value will be signaled.

sentinel value is true, the selector signal will

be set to the specified value and the input

NALU will be redirected to the correspond-

ing output position. The above-mentioned

selector signal is determined by a chaotic

sequence that proposed in our previous study

[3].

(d) Data embedding: Be fore transmitting the
bitstream to the cloud storage, it is necessary

to embed the privacy key informat ion in the

bitstream. The re lated data hiding process

can be found in our previous work [4].

2.2. Scrambling by Itself

In some situations, there is only one monitor

in the building. The second method is to scra m-

ble single video content by itself; the video

content can be recovered independently. The

ma in idea is that all the mac roblocks in the

identical picture will be switched to a new posi-

tion while the final bitstrea m is still for-

mat-co mp liant to a genera l H.264/A VC decoder.

The subset of macroblocks in a picture can be

listed below:

 MBrow = {mb1, mb2, ..., mbr-1}, where r is the
number of macroblocks in a row.

 MBcolumn = {mbr, mb2r, ..., mb( c-1)×r}, where
c is the number of macroblocks in a column.

 MBrest = {mbk | r+1 ≤ k ≤ r× c - 1, and k ≠ nr,

n ∈N
+
}.

Let set X is the original macroblock position

and set Y is the scramb le mac roblock position.

Given a one-to-one and onto function f that

assigns the set X to the set Y. Every ele ment of

Y is the image of unique element of X.

Due to the dependence on each adjacent

mac roblock, a huge drift error may be caused by

switching macroblock position. However, there

is a problem in switching macrobloc k position

directly. Because coeff_token is encoded by the

look-up table, switching macroblock position

directly may cause a violation of the standard

format. We need to re-encode the coeff_tok en to

ensure that it can be found in the look-up table.

3. Results and Discussion

The proposed video encryption scheme has

been imp le mented in a video stream platform

that is based on the Joint Model.

Proceedings of Engineering and Technology Innovation , vol. 3, 2016, pp. 19 - 21

3.1. Multi-way Videos Scrambling

(a) encrypted Hall monitor

(b) encrypted Container

Fig. 2 The encrypted video, frame #149, P picture

Fig. 2 illustrates a visual co mparison of

original videos and scramb led videos in the

8-way ca me ra environment. As can be seen, the

proposed scheme p rovides a high scra mb ling

effect that protects the video content from un-

authorized access.

3.2. Single Video Scrambling

(a) encrypted Hall monitor

(b) encrypted Container

Fig. 3 The encrypted video, frame #149, P picture

Fig. 3 de monstrates the scramb ling e ffect

which is quite d ifferent fro m the first method

and no video content is leak out. This method

provides a better scrambling effect.

4. Conclusions

In this paper, we propose two encryption

schemes that provide quite high scra mb le e ffect

for the video content in different scenario. It is a

real-t ime application since only simp ly parsing

and re-encoding in the co mpressed domain a re

needed. Although our approaches are applied

after the co mpression, we can s till p reserve

format-co mpliance. Future works a im to seam-

lessly combine two methods to achieve more

scrambling effects.

References

[1] Z. Shahid, M. Chau mont, and W. Puech,
“Fast protection of H.264/A VC by selective

encryption of CAVLC and CABA C for I

and P fra mes,” IEEE T rans. Circuits Syst.

Video Technol., vol. 21, no. 5, pp. 565-576,

March 2011.

[2] Y. Wang, M. O'Neill, and F. Kurugollu, “A
tunable encryption scheme and analysis of

fast selective encryption for CA VLC and

CABA C in H.264/A VC,” IEEE T rans.

Circuits Syst. Video Technol., vol. 23, no. 9,

pp. 1476-1490, February 2013.

[3] Y. T. Chang and Y. C. Lin, “The scra mb ling
cryptography imp le mented with chaotic

sequence trigger optical switch algorith m in

WDM passive optical network,” Proc. IEEE

Int. Ca rnahan Conference on Security
Technology, September 2015.

[4] W. S. Wang and Y. C. Lin, “A tunable data
hiding scheme for CA BAC in H.264/A VC

video streams,” Proc. IEEE Int. Sy mposium
on Ne xt-Generat ion Electronics, May 2015.