Deduplication Analysis of Products in Digital Marketing

ABSTRACT


INTRODUCTION
The extensive use of image editing software has induced an ascent interest in systems able to distinguish original images from tampered images and to create the validity of digital photographs. By enabling methods in existing forensic literature, forensic investigators can not only classify the possession environment of digital content, but also sense the processing history that the content has gone through after achievement. Today digital marketing has taken a new route in the industrial field. Now-a-days almost every camera is equipped with digital camera. It consist of various image editing software like Corel draw, Photoshop, Firework, etc. The introduction of these helps designed to enhance and improve the quality of the image. However some people used these tools to create tampered pictures and spread viral throughout the social media. Forensic analysis helps us to stand as an evidence for law enforcement.

RELATED WORK
Nowadays, in social media, we see a lot of faked images. These were created by using Forensic techniques. We are using tampering detection method [1], which is able to sense the active attempt to compromise the data integrity. Tampering detection based on image hashing to analyze and gather information about modified image. The modified images and the original images were segmented by image segmentation method. It is the process of partitioning the images. The segmented images were then matched with each other by image alignment technique. Forensic hash technique acts as evidence to extract information about the forensic images [5]. From the resulted forensic image the relationship among all the segments is ranked with the help of a manifold ranking technique. Gaussian Filter method was used by forgers because it eliminates noise and smooth transition [3]. It was also known as Gaussian Blur used to blur the images. It uses  [4] is the process of copying a region of an image which is scaled before pasting to some other location in that image. A novel method of the JPEG Ghost detection method was used to make judgment whether it is original or tampered [5]. It performs format based image forensic approach. Extracting a finger print of digital camera can be performed by pixel to pixel, non-uniformity, since every picture has an overlaid weak-noise. Knowing the value of digital camera we can identify whether it is source camera. Photo Response Non Uniformity (PRNU) [6] plunk as a reliable technique in fingerprint extraction of digital camera. It is important to find out the limit of capability of forensic by using an information theoretical framework. Forensic ability [7] is used to collect the maximum information that can be obtained from the action by performing some operation. The mutual information obtained having similarity in their concepts such as features and operations [2].
By using Mutual information the forensic ability is measured. The result is used to find out the Error Probability rate. In image processing, the units inherit the images in raster bitmap [8]. In this paper an efficient method is used to identify the previous JPEG compressions. A novel technique is used to find out tamper detection and tamper localization [1]. Image hashing algorithm was used. The tampering was identified in the small regions, corners, an exact location of the tampered image based on this the algorithm is applied using ring partition, Non Negative Matrix Factorization (NMF). The particular tampered region is detected using Image Hash Algorithm.
The variation of JPEG exploitation is used for image authentication. From a JPEG image a camera signature is extracted and it contains the information about quantization tables, Huffman codes, thumbnails etc [8]. 773 different cameras and cell phones were used across 1.3 million images. 62% signature of an image is unique by the single camera and 80% was shared by three or more cameras. The perceptual image hashing maps an image appeared to a human eye to a fixed length string. Its applications are image indexing, authentication and watermarking [10]. In this paper a general framework is used by feature points which should be largely invariant under distortion. An iterative feature detector is used to satisfy this property. In digital image forensic the JPEG [8] plays a vital role. The study of an image forensic is termed as JPEG error analysis. The main errors include quantization, rounding and truncation is obtained in JPEG. Single or double compression of an image is theoretically analyzed.

SYSTEM DESIGN
Forensic Data Analysis is the study of Digital Forensic. It helps to create tampered image which is stored in a structured database and to examine the fraudulent activities. Digital Camera captures the image of an object. Pixel analysis technique calculates the value of the two different images stored in database. Orthogonal transformation was carried out between these images. Digital value is calculated and creates a histogram to examine the difference between these images, show in Figure 1.

Pixel Analysis
The pixel is a basic unit in a computer display or an image. It takes as a logical unit rather than as a physical unit. A pixel is a sample of an image more samples are grouped together to produce a digital image. This tool helps us to drag a spatial position at a particular point. In this technique Color Based Segmentation using K-Means Clustering [13] method is performed. Reads the image from the database and convert the image from RGB color to Space portion using CIELAB. Which help us to differentiate pixel based on the color and partition it. Object in the cluster was performed to group the pixel. Blue Nuclei are the L value in CIELAB which segment the pixel color as light blue and dark blue. A and B represent the color component to segment the images. Various colors are tracked from the image and grouped together.  Purchased PDT image Object in cluster Blue nuclei Green object Red object 2. Color based segmentation using clustering for delivered image.

Principal Component Analysis
It is a geometric method for orthogonal transformation. To convert a set of interrelated variables to a set of uncorrelated variables is also known as principal components. PCA is sensitive to the relative scaling of the original variables. It is a statistical procedure used as a tool for data analysis and for making procedural models. The main purpose of PCA is to analyze a data and to identify patterns. In this method captured an image and the tampered image was converted to a grayscale image and then recovered the other image. Recovered images and the original image produce the difference between them in values of digital number and the elapsed time taken to complete the operation. From Figure3, we can see the elapsed time and digital number to distinguish between the object.

Least-Square Support Vector Machine
Least squares support vector machines (LS-SVM) is a set of related supervised learning techniques that examine data and identify patterns, which is used for classification and regression analysis. It is the fastest technique. It was based on the simple iterative approach. For example, it has been used to classify a dataset with 2 million points and ten features in only 34 minutes on a 400 MHz Pentium--II. Optimization tools in SVM are used for solving machine learning problem. In this method, value has been plotted to two different labels. The graph is drawn to relate these points.

Forensic Data Analysis
Forensic Data Analysis (FDA) is a study of Digital forensics. It examines structured data with consider to incidents of financial crime. The aim is to see and investigate patterns of fraudulent activities. Data from function systems or from their underlying databases is referred to as planned data. Analysis of large volume of data is handles by analysing team. The data copies of the images are aligned in separate database to prevent original images. To analyze large structure data with intense of detecting financial crime it takes three expertises from the analysing team.

SYSTEM ARCHITECTURE
On plan premises, we have various hardware set up along with software tools. Some of the tools are Microcontroller, Digital Camera, LED, Buzzer.
A microcontroller is a mini computer with Integrated circuits. At first it was programmed in assembly language later, it also supports high level programming languages like C, Python, JavaScript. Nowadays an embedded system software is used to code microcontroller.

Digital Camera
Digital Camera is used to take pictures and to store it on a computer or memory. Now, almost every phone is equipped with a digital camera. The Resolution of an image can be captured when light falls through the image.

Light Emitting Diode (LED)
LED is a semiconductor device produce light when Electric current passes through it. Various applications in which LED is performing is indicator light, LCD panel backlighting, fiber optic data transmission, remote control.

Buzzer
It is an audio signaling device which is a mechanical or piezoelectric. Uses of beeper which is also known as the buzzer is in alarm timings, timer, and conformation of input like mouse clicks or keyboard. The original or tampered image is detected when the led or buzzer glow by forensic technique.

Serial Adapter
USB adapter is a type of protocol converter which is utilized for converting USB data signals to and from other communications standards. Commonly, USB adaptors are acclimated to convert USB data to standard serial port data and vice versa.
Most commonly the USB data signals are converted to either RS 232, RS 485, RS 422 or TTL serial data. The older serial RS 423 protocol is infrequently utilized anymore, so USB to RS 423 adapters are less prevalent. Figure 7. USB to Serial Adaptor USB adapter is a type of protocol converter which is utilized for converting USB data signals to and from other communications standards. Commonly, USB adaptors are acclimated to convert USB data to standard serial port data and vice versa.
Most commonly the USB data signals are converted to either RS 232, RS 485, RS 422 or TTL serial data. The older serial RS 423 protocol is infrequently utilized anymore, so USB to RS 423 adapters are less prevalent.

5.
EXPERIMENTAL ANALYSIS X and Y are matrices holding the training input and training output. The -th data point is epresented by the -th row X(i,:) and Y(i,:). gam is the regularization parameter: for gam low minimizing of the complexity of the model is emphasized, for gam high, good fitting of the training data points is stressed. kernel_par is the parameter of the kernel; in the common case of an RBF kernel, a large sig2 indicates a stronger smoothing.
The kernel_type indicates the function that is called to compute the kernel value (by default RBF_kernel). The data copies of the images are aligned in a separate database to prevent original images. To analyze large structure data with intense of detecting financial crime, it takes three expertise from the analyzing time.
Scattered Function Estimation Using LSVM Data Points (Blue) And Estimation (Red) Figure 8. Scattered function estimation using LSVM for the tab Figure 9. Scattered function estimation using LSVM for the phone It performs histogram comparison between two images to calculate the difference between both. To calculate its difference it has to read two images and convert images to type double (range from from 0 to 1 instead of from 0 to 255). Reduce three channels [ RGB ] to one channel [ grayscale ] and then calculate the Normalized Histogram of Image 1 and Image 2.

CONCLUSION
Nowadays, in digital marketing has prone with tamper images. Our Proposed method is used to capture the image by a digital camera and to compare it with the structure database in the system to analyze its authenticity of an image. Tampered images are detected with the help of forensic data analysis technique. CIELAB was helpful in calculating the pixel value of an image. Principal Component Analysis is used to classify the Regression Analysis. Support Vector Machine classifies the relation between the images. Our future work is to use these techniques in various types of object. We can also investigate the global level performance of an object.