The paper proposes the technology of control of integrity and legitimacy of information object usage with the Look Up Table-oriented (LUToriented) architecture. The technology is based on embedding the self-recovery digital watermark into information objects of such kind. The technology is the composition of approaches to forming the self-recovery digital watermarks in the passive multimedia containers, and approaches to embedding the extra information into LUT-containers. The process of embedding the extra information occurs with the help of classification of container elements and their purposeful modification within the set of the formed classes. The procedure of immediate embedding the data at the level of LUT-container elementary parts includes the value inversion of current processed LUT unit and propagation of the inversion around all the inputs of LUT units connected to the current unit output. The description of practical realization of the proposed technology is represented.
DWM technologies are based on steganographical technique, with the help of
which the fact of DWM presence in an information object (DWM container) is
hidden. At the same time DWM can be read in the container if someone has a
stegokey possessing the set of access rules to DWM elements [
The control of information object integrity by DWM is based on embedding some
control data unit allowing to analyze the object integrity in an information object.
Hash sum calculated with the help of the definite hash-function is most frequently
used as such kind of data unit [
The control of information object usage legitimacy by DWM is based on
possibility of a copyright owner to embed DWM having the copyright owner
identifying data into the object. So it is only a copyright owner who possesses a
stegokey can check the DWM presence and its contents in the information object [
In the present paper the author proposes the information technology of
selfrecovery DWM usage in active hardware containers based on LUT-oriented
architecture (further LUT-containers). We can refer, for example, Field
Programmable Gate Array microchips (FPGA microchips) [
Self-recovering DWMs have been developed and used in the field of control of
integrity and legitimacy of usage for passive multimedia containers, e.g. graphical,
audio and video files. However it is necessary to control not only the information
objects of such kind by DWMs. But the task of self-recovering DWM usage in active
containers performing some calculating or controlling function is not realized at the
moment. In the articles [
The information technology considered in the paper is a formal process of usage of the proposed and already existing techniques as well as the means of information processing, which provides DWM forming in LUT-container space.
The proposed information technology uses a set of Fridrich-Goljan-Du method
(further Fridrich method) approaches [
The methods proposed in [
Fridrich method [
Function of discrimination is used in Fridrich method on order to classify the groups of elementary container parts by including them to the classes of regular, singular and unused groups. To determine the function value in each of the container groups we are to calculate the sum of pairwise subtractions with overlapping for elementary container parts included in a group.
Flipping-function is used by Fridrich method to modify the groups according to
some rule possessing the characteristic of involution. In the proposed information
technology the LUT-container architecture peculiarities are taken into account and
this fact differs it from Fridrich method in the following features:
within the frames of the proposed technology in the course of DWM embedding
the modification of elementary parts exactly (LUT units) is performed but not the
groups of elementary parts of a container;
within the frames of the proposed technology the calculation of Flipping-function
F(x) unlike Fridrich method contains the LUT unit value inversion according to the
procedure mentioned in [
Certainly, the rules of moving the container from one class to another defined by Fridrich method are performed within the frame of the proposed approach as well: F(LUTR) = LUTS; F(LUTS) = LUTR; F(LUTU) = LUTU, (1) where LUTR, LUTS, LUTU – LUT units classified as R-unit, S-unit and U-unit, respectively; F() – Flipping-function.
The succession of actions in the proposed information technology in DWM embedding in LUT-container is as follows.
Stage 1. The movement about LUT-container units in the order determined by embedding path is realized. During this action each of the LUT units refers to the classes R-unit, S-unit or U-unit on the basis of the above mentioned classification rules. According to the results of classification the binary RS-vector is formed in the following way: if the next LUT unit is classified as R-unit then zero value is fixed in RS-vector; if the next LUT unit is classified as S-unit then one value is fixed in RSvector; U-unit values are not fixed in RS-vector.
Stage 2. The obtained RS-vector compression is performed with the help of some lossless compression algorithm. As a result of this the compressed vector RScom is formed. It is obvious that RScom vector length is less than RS-vector length. Difference of length of these vectors is denoted as ΔL.
Stage 3. DWM, which is a binary succession the length of which does not exceed ΔL is added to vector RScom by concatenating. Vector RS* obtained after concatenation is added with arbitrary binary values until it reaches the RS vector length. So: RS* = RScom . DWM . Add, (2) where “.” – operation of concatenation; DWM – watermark embedded into container; Add – optional addition of vector RS* to RS vector length.
Stage 4. The movement about LUT units of container in the order determined by embedding path is performed. In the course of this movement U-units are ignored, and within R-units and S-units a mutual transition of one into another with the help of Flipping-function is made according to the table.
The actions at the given stage leads to vector RS* embedding in a LUT-container. Vector RS* contains DWM and a compressed vector RS (vector RScom) version in accordance with equation (2). Thus except for a DWM itself the information necessary for the original container value recovery is embedded in a container. The rules represented in table 1 describe the actions providing the vector RS* embedding in a LUT-container. If the corresponding current bits of vectors RS and RS* coincide then no modification for the current LUT unit is produced. In the case of mismatch of these bits the current LUT unit transition in the class corresponding to the current bit value of vector RS* is performed.
Taking into account that the Flipping-function impact on LUT units occurs
according to the unit inversion rules represented in [
Let us consider the example of described stages of DWM embedding in container technology. In table 2 the hexadecimal values of codes of twenty LUT units (i = 1…20), which are in some embedding path of a container are shown in the lines “Unit code”.
For each of these units a number of one values (n1) in the unit code bits are indicated. Depending on the proportion of amounts of one and zero values in the code the unit is classified as R, S or U unit according to the above mentioned rules. In accordance with the results of the classification a RS-vector is formed, in which zero values correspond to R-units and one values – to S-units. U-units do not participate in forming the RS-vector.
As a result the RS-vector takes the value RS = 011100001000001000 consisting of 18 bits (the information about two U-units was not included in RS-vector). Then the RS-vector is to be subjected to lossless compression. In the given example the simplest way of compression on the basis of Huffman method is used for illustration (in practice the more effective compression methods are to be used). For this purpose the triads of RS-vector bits are taken as elementary characters and their frequency of entering the vector is found. The triad “000” enters the vector twice, the triad “100” – once, the triad “011” – twice. As a result of Huffman method we obtain the following system of uneven prefix codes for the triads of RS-vector: “000” “11”, “100” “100”, “011” “101”, “001” “0”.
As a result of usage of the obtained codes we can have 12 bits vector RScom = 101100011011 instead of the initial triads in 18 bits RS-vector. RS and RScom vector lengths difference is ΔL = 6. This value expresses the maximum amount of DWM bits, which can be embedded in LUT units in the given example.
In table 3 the values of bits of the initial RS-vector and its compressed version RScom are shown. Let us consider the example of addition of the 6-bits DWM – DWM = 101010 to the vector RScom. As a result of concatenation of the vector RScom and DWM embedded in DWM container we obtain vector RS* having the similar length as to vector RS.
RS RScom DWM RS* 0 1 1 1 0 0 1 1 1 1 1 1
Then according to the rules specified in table 1 the LUT-container modification is performed. The aim of this modification is to adapt the LUT-unit classes lying in the embedding path to the value of vector RS* containing DWM. As we see from table 1 the LUT-unit movement from one class to another is only performed when the values of the corresponding bits of vectors RS and RS* do not coincide. The bits characterizing with the absence of coincidence of such kind are distinguished in table 3. For the LUT-unit codes corresponding to the distinguished bits a Flipping-function is applied and this leads to the movement of these units from one class to another.
In table 4 the values of LUT-unit codes after Flipping-function application to the distinguished units (units, for which the class is to be changed). As a result of these actions the RS-vector of units, which are in the embedding path takes a value of the vector RS* and DWM is embedded into the LUT-container.
Within the frame of the proposed technology the stego-key dedicated for data extraction consists of the following four components: key = (order, classification, RSrule, ΔL), (3) where order – the information defining the order of move around LUT units in the container (embedding path) for DWM embedding or extracting; classification – concrete definition of the principle of LUT-unit classification (unit is considered to be R-unit if the amount of parts in the unit code is more or less); RSrule – the rule of interpretation of RS, RScom, RS* vector bit values: for R-units a zero value and for Sunits one value are fixed in the RS-vector and vice versa; ΔL – the difference between the vector RS length and its compressed version RScom length.
The succession of actions of the proposed information technology in DWM extracting and recovering the initial value of LUT-container is as follows.
Stage 1. The movement about LUT-container units in the order specified by embedding path occurs. And along with this a binary vector RS’ is formed on the basis of LUT-unit classification like in the case of embedding the information.
Stage 2. The obtained vector RS’ structure corresponds to the one of vector RS* (2) formed in DWM embedding in container. The last ΔL bits of vector RS’ are DWM with the possible addition to the necessary vector length. At this stage the reading of this bits and DWM obtaining is performed.
Stage 3. The other vector bits are subjected to the procedure of decompression opposite to compression, which is performed at the stage of embedding the information. As a result a binary vector RSdecom is formed.
Stage 4. On the basis of information contained in vector RSdecom the recovery of original container value is performed. The movement about LUT-units of the container in the order specified by the embedding path and successive review of vector RSdecom values are performed for this. In the course of this movement the Uunits are ignored and within R-units and S-units a mutual transition is made by Flipping-function according to table 1.
For the example considered above DWM extraction can be described in the following way. As a result of analysis of LUT-unit codes lying in the embedding path the unit classification is performed and a binary vector RS’ = 101100011011101010 is formed. From the vector end the ΔL = 6 bits containing DWM “101010” are taking. The rest bits of the vector “101100011011” are subjected to decompression with the usage of the table making the match between the uneven prefix codes of compressed vector and the triads of decompressed vector bits. This leads to creation of binary vector RSdecom = 011100001000001000, which coincides with the container vector RS before DWM embedding in it. The given vector has the information for the original value recovery of LUT-container. According to the rules of table 1 the LUT-unit classes become the original ones and the container acquires the value it had before DWM embedding. 5
1. to show experimentally that LUT-containers, in which DWMs were embedded with the help of the proposed technology acquire their original value after DWM extraction; 2. to indicate the degree of change of the main characteristics of containers after DWM embedding in them. The main characteristics are considered to be the following ones: a) maximal clock signals frequency expressing the limits of processing speed within the frames of one family of target chips; b) energy consumption of the input-output system of a chip; c) energy consumption of the chip core; d) thermal dissipation of a chip.
Environment for making the experiments. The hardware-software means based on the chips FPGA Altera Cyclone II and CAD Altera Quartus II usage have been developed for the experimental research. The group of scripts performing the interaction with CAD Altera Quartus II for reading and recording the LUT-unit contents was organized in TCL language. The read data processing subsystem itself is realized in language C# within the frame of the platform .Net in accordance with the proposed technology. The estimation of design characteristics mentioned above were carried out by means of CAD Altera Quartus II Timing Analyzer (estimation of the limit processing speed) and Power Play (estimation of energy consumption and thermal dissipation). The states of container before DWM embedding and after DWM extraction were compared by bit-by-bit analysis of configuration files of the corresponding containers.
complexities and purposes. The hardware complexity of devices within the frames of the given projects varies from 1,2% to 65% resource volume (logic cells, RAM units) of target chip FPGA.
The process of making the experiments. The experiments for each of the researched project consist of the following stages: 1. estimation of the main characteristics (processing speed, energy consumption, thermal dissipation) for the device represented in the form of initial container; 2. embedding the random binary succession of DWM in container; 3. estimation of the main characteristics for a container having the embedded DWM; 4. DWM extraction from the filled container; 5. comparison of configuration files of the final and initial container states.
coincidence of configuration files of the initial containers and the ones obtained after DWM extraction for all of the 40 projects has been established. So due to the features of self-recovery after DWM extraction from containers they take the original form and their characteristics return to the initial values.
In the part of determining the degree of changes of the main container characteristics (after DWM embedding) the following results were obtained. We found that insignificant changes of container characteristics occurs as a result of DWM embedding. Along with this the dependence of this change upon hardware complexity and its structural peculiarities is extremely low. The difference between the changes of characteristics for devices occupying 1,2% and the ones occupying 65% resource volume of the target chip FPGA is hundredths of a percent. On average (for all of the researched 40 projects) the processing speed characteristic change is 0,18%, changes of energy consumption and thermal dissipation characteristics – 0,22%. As we see the change values are within the limits of error value of measurement means.
The technology offered in the paper does not change the mutual connections of
LUT units but does change the values of the specific units codes. It is accordingly of
interest to learn whether the changes of units codes are able to influence on the FPGA
project features under such conditions. Except for the previously considered
experiment another experimental research of the influence of the values of LUT units
codes on the basic features of FPGA projects has been made with the participation of
one of the authors of the given paper. The description of this research and its results
are shown in [
Pmax δ
Pmax δ
Pmax δ
Pmax δ
The experiment results shown in table 5 demonstrate that in maximum mass change of the number of one values correlation in LUT units: 1. the energy consumption of input output system for chips within a FPGA family is retained at the same level; 2. the chip core energy consumption changes insignificantly within one family mainly due to the dynamic component (maximum value of this change equals 0,46% for chips EP4CE30F29C6 of family Cyclone IV E); 3. the maximum clock frequency expressing the extreme processing speed of devices does not practically change within one FPGA family.
The results show that such important characteristics as productivity and energy consumption are not practically dependent upon the type of Software code used in FPGA-projects with the fixed hardware realization. So we can come to the conclusions that according to the technology proposed in the given paper the DWM embedding in a LUT-container does not substantially impact on the FPGA project characteristics mentioned above.
The information technology proposed in the paper allows to perform DWM embedding in containers with LUT-oriented architecture. The technology gives the possibility to recover the initial state of container after DWM extraction from it. The technology is based on the combination of DWM embedding method into LUTcontainers proposed in the given paper and the popular Fridrich method oriented at the passive multimedia containers. Unlike Fridrich method the proposed technology: uses less complicated Flipping-function taking into account the container peculiarities; does not include the discrimination function calculation in order to perform the element classification of container; performs information embedding at the level of the container elementary parts (LUT-units) but not at the level of groups of elementary parts.
The degree of the proposed technology effectiveness is of qualitative nature and is expressed in possibility to recover the initial LUT-container after DWM extraction, which was absent before. The proposed technology can be used in developing the hardware and software, which realize digital watermark embedding in computer and control devices created on the LUT-oriented element base (e.g. FPGA or programmable logic integrated circuits with the similar architecture). Such kind of embedding allows to control the device configuration integrity makes the substitution or corruption of configuration impossible, which is of vital importance for critical domains. In addition it allows to control the legitimacy of usage of project information and devices themselves at the different stages of CAD and life cycle: synthesized FPGA project, configuration file FPGA, operating device.