DIGITAL PROCESSING METHODS OF IMAGES AND SIGNALS IN ELECTROMAGNETIC INFILTRATION PROCESS

Transcription

1 Image Processing & Communication, vol. 16,no. 3-4, pp DIGITAL PROCESSING METHODS OF IMAGES AND SIGNALS IN ELECTROMAGNETIC INFILTRATION PROCESS IRENEUSZ KUBIAK Military Communication Institute, Zegrze, Poland, i.kubiak@wil.waw.pl Abstract. The article contains information about the capabilities of electromagnetic infiltration process in case of occurrence of strong interfering signals. As a methods supporting infiltration process used method of digital processing of signals and images in the form of histogram transformations, global and local thresholding of signal amplitudes and logical filters. The material presented in the article shows that risk can arise if the uncontrolled use of the computer. Risks that could decide our safety and security of our data. 1 Introduction In the era of powerful computing life, protection against electromagnetic leakage of information is becoming increasingly important. Despite lower energy demand of equipment which transforms classified information, increased data rates, the risk electromagnetic infiltration is not reduced. Lower levels of compromising emanations necessitate exploration of new methods to extract data from the noise and disturbance. The essence of these processes are the method of processing digital signals and obtained images. Manipulation of histograms, threshold amplitudes of the emission signal correlated with the signals classified or logical filters highlight the weakness of the security used at the source. The presence of strong interfering signals such as vertical and horizontal synchronization signals blocking measurement receivers, do not prevent the reproduction of classified information. Opportunities of the electromagnetic infiltration in situations of weak compromising emissions occurs are presented in the article. 2 Primary sources of compromising emanation The threat identification and recovery of information from emission signals revealing the most supported in the event of influence their effects on human visual stimulus. These types of signals are video signals coming from computer stations, which graphic lines are the main sources of unwanted emissions. In addition, video signals are repeated signals which facilitates the possibilities of unwanted recording. In the case of computer units in particular CPUs and monitors are secured. There are sophisticated methods using special shielding windows, filtering the power and signal circuits. However the transmission medium which are the cables and connection of the ca-

2 2 I.Kubiak bles to the monitor and CPU are weak elements. They in many cases constitutes a major threat to the disclosure of our data. This applies to both standard VGA and DVI. The use of cryptographic security of these cases does not bring the expected results. Transfer information from the graphics card to the monitor are in the overt form. The only protection is to use properly shielded cables, which can effectively reduce the level of radiated emissions revealing. 3 Methods of digital images and signals processing used in electromagnetic leakage of information 3.1 Introduction In situations where there are strong environmental disturbances or disturbances directly related to the work of "intercepted" (tested) devices, electromagnetic infiltration process in real time is difficult. The only way that could facilitate the reconstruct of classified information is registration of "suspicious" signals and than their digital transformation. Here helpful are methods of digital transformation of images and signals. The choice of methods should be remembered that the images are not typical photographs. Images created in the rasterization process from revealing emission signals have distortions arising from the transition of signals through the channel of leakage information. Such channel is a differential channel. This means that except the strong interfering signals and significant levels of noise only vertical and diagonal edges of the figures appearing in the original image are visible in the reconstruct images. In addition, the output of the receiver used in the infiltration process available the signal module which provides an additional obstacle to the process of reconstruct primary information. For digital processing of the recorded signals and images obtained from that signals, used three methods of computer analysis and processing images: histogram transformation: alignment; extension; thresholding: of the average; two-step for the two thresholds; non-linear filters-logical. Each method allows reconstruct data masked in the images. The operations of these methods rely on: to change the amplitude of the pixel through analyzing the amplitudes of neighboring pixels (local methods); analysis of the amplitudes of pixels throughout the image, definition of the thresholds and the comparison of them to value amplitudes of pixels (global methods); quantitative change pixels about known values of the amplitudes (global methods). 3.2 Histogram transformations Images obtained in the electromagnetic infiltration process have many shortcomings that make difficulties to identify the graphic elements contained the images. Most images are too dark, despite the use of the full dynamics sampling. It is result of the presence of points and continuous (sync signal) interference with significant value of amplitudes which blackout the images. Effective methods eliminating this type of distortion could be transform the image histogram. The use of these methods changes the quantitative distribution of pixel amplitudes compared with the original image. From range available transformation histograms (alignment, line extension, the extension of linear segments),

3 Image Processing & Communication, vol. 16,no. 3-4, pp Fig. 1: Image and its histogram obtained from compromising emanation for mode 640x480/60Hz of extending a selected range of the original histogram. Both operations enough emphasize graphic elements hidden in the image in presence of strong interference. The considerations made for a set of computer operating in two modes: 640x480/60Hz and 800x600/60Hz. After recording the compromising emissions were appropriate digital treatments. Images obtained with the emission signals without the use of digital treatments to improve quality, were shown on Fig.1 and Fig.2. Pixel amplitudes for the two considered cases are concentrated in the gray color scale corresponding to brightness from 20 to 30 (Fig.1) and from 15 to 40 (Fig.2). If we assume that p = (n, m) is the amplitude of the pixel with coordinates (n, m) (where n [0, n 1], N - the height of the image, m [0, m 1], M - width of the image), then for the cases shown in above figures: figure 1 - average value of brightness of image pixels p = 1 (N 1)(M 1) N 1 M 1 n=0 m=0 - variance of brightness image (contrast) p(n, m) = 23 (1) σ 2 = 1 (N 1)(M 1) N 1 M 1 n=0 m=0 (p(n, m) p) 2 = 144 (2) figure 2 - average value of brightness of image pixels Fig. 2: Image and its histogram obtained from compromising emanation for mode 800x600/60Hz not all successfully extracted data from the images. Practical tests shows the greatest efficiency of the extension of linear segments of the histogram and the modified method p = 1 (N 1)(M 1) N 1 M 1 n=0 m=0 - variance of brightness image (contrast) σ 2 = 1 (N 1)(M 1) N 1 M 1 n=0 m=0 p(n, m) = 23 (3) (p(n, m) p) 2 = 441 (4)

4 4 I.Kubiak Parameter σ 2 called the image contrast is defined as the variance of image brightness. Greater value should be informed about more clarity and readability of its components. The purpose of extracting hidden information contained in images which are shown in the Fig.1 and Fig.2, was carried out operation of extension histogram (extension of brightness) using linear segments and a modified method of extension of the selected interval histogram of the original image. Normal extension of the histogram, where the pixel colors occupy almost the whole range of gray scale does not produce the expected results. a) extension of histogram with linear segments Extension the histogram with linear segments allows for any change in the brightness of image pixels according to the inclination of the linear segments the axis of abscissa (Fig.3). In this case the new values of the pixels p w (n, m) can be concentration or extension to gray scale P w in relation to the original pixel values p(n, m) contained in the scale P. Depending on the adopted values p 1, p 2, p w1 and p w2, we obtain the variety of inclinations linear transform sections which influences the change of the brightness of transformed pixels. Let the new brightness values of pixels will be described the equation: p w (n, m) = Fig. 3: Graph transformation functions for linear segments b 1 = 0 a 2 = p w2 p w1 p 2 p 1 b 2 = p w1 a 2 p 1 a 3 = p wmax p w2 p max p 2 b 3 = p w2 a 3 p 2 Selecting appropriate values p 1, p 2, p w1 and p w2, most often by trial-and-error method, we can extract a number of important graphic information from the transformed images which are important for the electromagnetic infiltration. For the cases shows in Fig.1 and Fig.2, the parameter values adopted for the transform functions of linear sections are shown in Tab.1. = a 1 p(n, m) + b 1 for p min p(n, m) p 1 a 2 p(n, m) + b 2 for p 1 p(n, m) < p 2 a 3 p(n, m) + b 3 for p 2 p(n, m) p max (5) Tab. 1: The parameters of transform function for linear segments Parameter Figure 1 Figure 2 p p p w p w where: a 1 = p w1 p 1 Influence of histogram transformation on the image quality, particularly on the visibility of hidden data in

5 5 Image Processing & Communication, vol. 16,no. 3-4, pp.1-8 the original images is enormous (Fig.4 and Fig.5). Previously unreadable data have become not only visible but also readable. Natural extension of the histogram for images obtained from the compromising emanations, does not improve the image quality for reading data. Fig. 5: The image and its histogram obtained by the extension of the histogram (Fig.2) using linear segments searched data hidden in an image do not different from the surrounding background. It has connection between the amplitudes of pixels carrying the information and the values of the amplitudes of the background pixels. This Fig. 4: The image and its histogram obtained by the exphenomenon is noticeable on the histograms shown in tension of the histogram (Fig.1) using linear segments Fig.1 and Fig.2. Therefore, the operation of the histogram extension should be submit only a fragment of the histab. 2: The parameters of the linear sections of the extension histogram Parameter Figure 4 Figure 5 average value of brightness of image pixels variance of brightness image (contrast) togram. The extension operation of the brightness scope (the histogram extension) can be expressed by the relationship: pw (n, m) = p(n, m) p1 (pmax pmin ) + pmin (6) p2 p1 b) extension of a selected range of the original image histogram where: Another method greatly improves the image quality is p1 - minimum value of amplitude of image pixel for the the extension of the original image histogram. How- considered fragment of the histogram that is lower limit ever, it is not typical operation of the histogram exten- of the histogram fragment; sion on whole the available range, that is from 0 to 255 p2 - maximum value of amplitude of image pixel for the (for grayscale). For images vary often is important part considered fragment of the histogram that is upper limit of the histogram. This follows from the fact that the of the histogram fragment;

6 6 I.Kubiak p(n, m) - value of amplitude of image pixel; p min - minimum value of the wide range of brightness, in the particular case equals to 0; p max - maximum value of the wide range of brightness in a particular case equals to 255. Taking p min and p max accordingly the values 0 and 255, the equation (6) takes the form: p w (n, m) = p(n, m) p 1 p 2 p (7) Tab. 3: The threshold values p 1 and p 2 for the operations of the histogram extension and parameters obtained images Parameter Figure 6 Figure 7 p p average value of brightness of image pixels variance of brightness image (contrast) Analyzing the above equation it should be noted that: p w (n, m) < 0 for p(n, m) < p 1 (8) p w (n, m) > 255 for p(n, m) > p 2 (9) that is the limits of the gray scale are exceeded, which are within the limits. It is necessary to introduce some modifications of the equation (3), by adopting appropriate conditions, allowing to acceptance by the output image pixels value of the amplitudes within the scope of that scale. Let therefore: p w (n, m) = where: K 1 p 1 p(n, m) p 2, p(n,m) p 1 p 2 p for K 1 p(n, m) for K 2 (10) Fig. 6: The image and its histogram obtained by the extension of the histogram fragment (Fig.1) 3.3 Threshold values of pixel amplitudes K 2 p(n, m) < p 1 lub p(n, m) > p 2. Using the operation of the histogram extension described by the relation (6), for images from Fig.1 and Fig.2 for given threshold values p 1 and p 2 (Tab.3), obtained images presented on Fig.6 and Fig.7. Threshold values of the amplitudes of image pixels is one of the fastest and most effective methods to expose hidden data in the background image. This is most useful for cases when the outlines of shapes of sought elements are noticeable. For need of analysis connection with the compromising emanation used in two methods. The first is related to the appointment of the average value of amplitude pixels:

7 Image Processing & Communication, vol. 16,no. 3-4, pp Fig. 8: The image obtained by thresholding the value of the average value of amplitude pixels for the image from Fig.1 (p = 23) Fig. 7: The image and its histogram obtained by the extension of the histogram fragment (Fig.2) 1 N 1 M X X 1 p(n, m) p = (N 1) (M 1) n=0 m=0 (11) and adopt it as the threshold, according to the following relationship: pw (n, m) = 255 for pw (n, m) p Fig. 9: The image obtained by thresholding the value of the average value of amplitude pixels for the image from (12) Fig.2 (p = 23 ) 0 for pw (n, m) < p Thresholding method of average value is an operation that threshold (Amax and Amin ) by trial-and-error method. does not require visual analysis of the amplitude pixels, Too low value of Amax, can decrease the amplitudes of which could be used to determine appropriate values of information-bearing pixels to a minimum, by which the thresholds (Fig.8 and Fig.9). requested data will be lost. A similar situation may occur when determining parameter Amin. Too low value of Another alternative method of thresholding of the Amin, can increase amplitudes of the background image amplitude pixels is minimization of the amplitude pixels to the maximum. (maximum threshold Amax ) and maximization (mini- Thresholding operation with threshold Amax and Amin is implemented in two stages. mum threshold Amin ) these values in case fulfillment The first stage is set threshold Amax. The amplitude appropriate the set conditions. This approach however, values of the pixels larger than Amax are reduced to the forces the image analysis and search for the optimal minimum pixel amplitude of image (or 0). In the second

8 8 I.Kubiak stage is determined threshold Amin. The amplitude values of the pixels larger than Amin are increased to a maximum pixel amplitude of image (or 255). Write the mathematical operations described as follows: a) first stage pw (n, m) = pmin for p(n, m) Amax p(n, m) for p(n, m) < Amax (13) Fig. 10: The image obtained by two-stage thresholding of where pmin is the minimum value of the amplitude pixel amplitudes for the image from Fig.1 pixel of image undergoing transformation (in the particular case pmin = 0). b) second stage pw (n, m) = 255 for p(n, m) Amin p(n, m) for p(n, m) < Amin (14) A two-stage thresholding for values Amax and Amin makes a significant way may be eliminated from the image noise for large values of the amplitude pixels. The ef- Fig. 11: The image obtained by two-stage thresholding of pixel amplitudes for the image from Fig.2 fect of two-stage thresholding for value thresholds A max and Amin contained in Tab.4 are shown in Fig.10 and Fig.11. filters can be compared with thresholding of amplitudes. In the case of logical filters is required, as the name says, Tab. 4: Value of parameters Amax and Amin Parameter Figure 10 Figure 11 Amax Amin the fulfillment of a condition basing on values of the amplitude pixels be presenting in a neighbourhood of a considered pixel. For the analysis of images obtained from the signals of compromising emanation the logic horizontal filter was used. The action of this filter is based 3.4 Nonlinear filters - logical on an analysis of the amplitude pixels located on the left and right side of the analyzed pixel p(n, m). If n is the Family of nonlinear filters is very different. But from the number of the next line of image and m - the number point of view of the electromagnetic leakage information of the next column of image, then specify the value of the most interesting nonlinear filters are logical filters. the amplitude pixel pw (n, m) is done by analyzing and Performed conversion of amplitude pixels by this type fulfillment of the relevant condition by the amplitude

9 Image Processing & Communication, vol. 16,no. 3-4, pp pixels p(n, m 1) and p(n, m + 1). For purposes of analysis of electromagnetic leakage information assumed the following rule: pw (n, m) = 255 for K3 pw (n, m) for K4 (15) where: K3 p(n, m 1) p(n, m + 1) > B, K4 p(n, m 1) p(n, m + 1) B. and B is a given and fixed by trial-and-error method the threshold for a particular image. In order to obtain the best effect can be several times the use of filtration process. Fig. 13: The images obtained by applying a one-filtration operation using horizontal logical filter for the image from Fig.2 Fig. 12: The images obtained by applying a one-filtration operation using horizontal logical filter for the image from Fig.1 There are other logical constructs filters. For example, Fig. 14: The images obtained by applying four times the a vertical or cross filter. However, their analysis showed filtration operation using horizontal logical filter for the less usefulness in the recovery process of data hidden in image from Fig.1 the images obtained from the compromising emanation than the filter described in equation (15). The results of were shown on Fig.12-Fig.15. The values of the parameuse of horizontal filter for images from Fig.1 and Fig.2 ter B for these cases were given in Tab.5.

10 10 I.Kubiak of histograms, threshold of the average value and logical filters in a large extent support the possibilities to detect of hidden data in the images. However, the use of each of the methods presented in the article requires experience in the field of detect information from signals correlated with the classified information. Sometimes the determination of basic parameters of the methods used must be done by trial-and-error method, which requires more time. However, the final results obtained largely compensate lost time. Fig. 15: The images obtained by applying five times the filtration operation using horizontal logical filter for the image from Fig.2 Tab. 5: The values of parameter B for horizontal filter and the parameters of obtained images Parameter Fig. 12 Fig. 14 Fig. 13 Fig. 15 B 3 6 average value of brightness of image pixels variance of brightness image (contrast) Fig. 16: The image obtained by use of a typical equalization histogram of image from Fig.2 A characteristic feature of the method of image trans- 4 Summary formations discussed in the article is the possibility of their independent use. This means that the obtained ef- Presented in this paper the methods of digital image pro- fect of one method is so satisfactory that does not require cessing are the main base transformations of images used a different method the next stage of improving readabilin the process of reproducing information from the com- ity. In the article was mentioned that the use of a typipromising emanations. Next typical operation of sum- cal equalization histogram does not bring the expected remation of images (required to record a signal of length sults. Additional procedures are required to increase readpossible to obtain an appropriate number of images), me- ability. For example on Fig.16 was shown the effect of dian filtering or a low and high passfilter, modifications the operation of a typical equalization histogram of image

11 Image Processing & Communication, vol. 16,no. 3-4, pp by the equation (14) (Fig.17). Presented in this article discussion about the applications of digital image processing methods also have reference to other modes of computer monitors (such as 1280x1024, 1920x1200). References [1] Kubiak I., Przybysz A., Musial S., Grzesiak K., 2012, Raster Generator in the Electromagnetic Infiltration Process, Military University of Technology [2] Richard G. Lyons, 2004, Understanding Digital Signal Processing (2nd Edition) Fig. 17: The image obtained by use of a typical equalization histogram of image from Fig.2 and thresholding described by the equation (14) for the value of threshold equal 105 from Fig.2. Its readability is so weak that it is necessary to apply next transformations such as thresholding described [3] Zielinski T., 2009, Digital Signal Processing, Communication and Transport Publisher [4] Deok J. Park, Kwon M. Nam, 1995, Multiresolution Edge Detection Techniques, Patern Recognition, Vol. 28 [5] Pratt K., 1978, Digital Image Processing, Wiley and Sons, New York