

Articles
by
Guang Sun 
Total Records (
7 ) for
Guang Sun 





Hengfu Yang
,
Xingming Sun
and
Guang Sun


A novel semifragile watermarking scheme in image spatial domain is proposed. Each watermark bit is duplicated and embedded into subblocks of the host image by adaptive Least Significant Bit (LSB) substitution. The adaptive LSB substitution fully exploits the Human Visual System (HVS) masking characteristics, which ensures high visual quality of watermarked image. The watermark for authentication is extracted by taking a majority vote on the extracted bits. To differentiate attack types effectively, a classification rule for image authentication was developed. Experimental results show that the proposed scheme has good transparency and robustness against admissible signal operations, while it is sensitive to malicious attacks such as heavy noise addition, rotation by large angle, cutting and pasting. In addition, the scheme can localize the tampered region precisely. 




Zuwei Tian
and
Guang Sun


In this paper, we analyzed the ifconversion algorithms based on IA64’s predicated execution in GCC,
and pointed out some existing problems of GCC ifconversion. And then an amended algorithm is proposed
which applies ifconversion selectively by considering multiaspects of factors, such as critical path length,
resource usage height, misprediction rates, misprediction cost and instruction number. The experiment results
show that the proposed algorithm can improve the effect of ifconversion and promote the performance of
programs. 




Guang Sun
and
Xingming Sun


This study presented a method for constructing more accurate condensed cochange graph. By using random walk, all the unchanged artifacts can be merged completely on the condition that the changed artifacts greater than 50%. An extended GN clustering approach is applied to find a good partition of the software condensed cochange graph. The experiment compares the watermark extracting results embedding whole software and embedding the biggest cluster codes by SANDMARK with ten software techniques. The experiment results show that if an attacker copies the biggest cluster codes, only three software watermarking approaches can work. If the watermarks are embedded in the biggest cluster codes, seven watermarking techniques can work. 




Chaoqun Liu
,
Yang Zhou
,
Yunhua Xiao
and
Guang Sun


According to the characteristics of hash function, the oneway hash encryption algorithm based on high diophantine equation (RSH) is proposed. RSH not only can be used for password encryption, but also can be used for Data integrity check and the digital signature of message digest. These algorithms hash the arbitrary length message into 128 bits, then make the multiiterations by high diophantine equation, finally produce the numeral string of 128 bits. In the conversion process, because we do not know the high indefinite equation iterative power law, so the resulting string of numbers is very reliable and safe. 




Lijun Tian
,
Chaoqun Liu
and
Guang Sun


This study presented an Undirected Network Construction Method. UNCM initializes a network with one vertex, zero edge and adds vertices one by one. Every time a vertex was added, a sequence of random variables ξ_{n,r} was given to determine the presence or absence of edges. ξ_{n,r }has a (01) distribution with probability p_{n,r} independently. There are three types of probability sequence p_{n,r}: specified, completely random and constrained random. By using a specified sequence, regular networks or other specific network can be constructed and by completely random sequence, some types of networks such as Poisson random graphs can be constructed. To construct more complex networks, a constrained random sequence is needed. As an example, configuration model was discussed. Theoretical analysis shows that any types of undirected networks can be constructed by UNCM, which means the research on how to construct such networks can be converted to the research on the characteristics of probability sequences p_{n,r}. 




Jianquan Xie
,
Xiaoping Fan
,
Hui Peng
,
Qing Xie
and
Guang Sun


Imperceptibility, robustness, security and hiding capacity
are important indexes in evaluating information hiding algorithms. Recently,
in most case, only imperceptibility and robustness were considered rather than
the security ability. The later one is highly required to guarantee secret communication.
With fully consideration of this requirement, one kind of hiding information
algorithm is proposed based on run length, the main idea of which is to separate
the image into several binary images first. Then according to the parities of
the run length of the binary images, one bit of information can be successfully
embedded while no more than one pixel that located along blackwhite boundary
of the binary image is modified. The blind extraction of the hiding information
can be achieved. The algorithm neither obviously changes of distribution property
of 0 and 1 in image’s low plane, nor reduces the number of long run length. Hence
it can successfully defend various detections that against Least Significant
Bit (LSB) algorithm and improved LSB algorithm and it can also be applied in
secret communication and other situations which require high security and large
capacity. 




Hangjun Zhou
,
Zhongli Liu
,
Sha Fu
and
Guang Sun


In a largescale peertopeer DVE (Distributed Virtual Environment) system running on the WAN (Wide Area Network) without the central control node, it is one of the most challenging problems to preserve the causality of all the simulation events on each node with asynchronous clock. However, the most existing asynchronous causality control methods seem to be inefficient to resolve the problem for applying either too closely coupled causal control information with system scale to achieve the realtime preservation or too little control information in a message to maintain the correct causality. In this paper, a novel causal order control approach with the deduction of the minimal calculable moment and the distributed preservation algorithm is proposed to achieve the correct causeeffect relationship among all the simulation events on each node. The experimental results demonstrate that comparing with the representative existing methods the proposed approach can cost lower transmission and calculation overhead with proper amount of causal control information in a message and ensure the realtime preservation of causality with asynchronous clocks. 





