Scribd Logo
Upload

Welcome to Scribd!

  • 23 views

    Seminar

    Uploaded by

    sreya.sreejith2211
    The document proposes using a convolutional neural network (CNN) to detect image forgeries, describing a system that uses a CNN model trained on manipulated and authentic image datasets to classify input images, extracting features through convolution, pooling, and activation functions to determine if an image is authentic or forged.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    Seminar

    Uploaded by

    sreya.sreejith2211
    23 views34 pages
    The document proposes using a convolutional neural network (CNN) to detect image forgeries, describing a system that uses a CNN model trained on manipulated and authentic image datasets to classify input images, extracting features through convolution, pooling, and activation functions to determine if an image is authentic or forged.

    Original Title

    Seminar.

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document proposes using a convolutional neural network (CNN) to detect image forgeries, describing a system that uses a CNN model trained on manipulated and authentic image datasets to classify input images, extracting features through convolution, pooling, and activation functions to determine if an image is authentic or forged.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    23 views34 pages

    Seminar

    Uploaded by

    sreya.sreejith2211
    The document proposes using a convolutional neural network (CNN) to detect image forgeries, describing a system that uses a CNN model trained on manipulated and authentic image datasets to classify input images, extracting features through convolution, pooling, and activation functions to determine if an image is authentic or forged.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 34

    IMAGE FORGERY DETECTION USING

    CNN

    Presentation for CSQ 413 Seminar


    SREYA T
    TKR20CS074

    Guide:Mrs Shana Musthafa A P V, Assistant Professor


    in CSE
    Department of Computer Science and Engineering
    College of Engineering Trikaripur, Cheemeni

    1 10/17/2023
    SREYA T(TKR20CS074)
    CONTENT

     INTRODUCTION
     PROBLEM STATEMENT
     MOTIVATION
     LITERATURE SURVEY
     WHAT IS NEURAL NETWORK?
     PROPOSED METHOD
     SYSTEM ARCHITECTURE
     DATASET DETAILS
     WORKING OF CNN
     RESULT
     CONCLUSION
     REFERENCES

    2 SREYA T(TKR20CS074) 16/10/2023


    INTRODUCTION
     Image forgery refers to the act of altering or manipulating
    digital images to deceive or mislead viewers.

    10/17/2023
    3 SREYA T(TKR20CS074)
    IMAGE SLICING

    COPY MOVE FORGERY

    4 SREYA T(TKR20CS074) 10/17/2023


    IMAGE RETOUCHING

    5 10/17/2023
    SREYA T(TKR20CS074)
    PROBLEM STATEMENT

     With the widespread use of image editing tools, there is a growing


    concern about image forgery and the spread of misinformation through
    manipulated images.

     CNN-based image forgery detection system offers a robust and


    effective solution to the problem of image manipulation and forgery in
    today's visual media landscape.

    10/17/2023
    6 SREYA T(TKR20CS074)
    MOTIVATION

     The increasing availability of cameras and the popularity of


    photography.

     Outperformed the existing methods for image forgery detection and


    demonstrated system potential for various applications, including
    forensics, security, and digital image analysis.

    7 SREYA T(TKR20CS074) 10/17/2023


    LITERATURE SURVEY
    Sl TITLE AUTHOR TECHNIQUES HIGHLIGHT LIMITATIONS
    no (YEAR) USED

    1. An efficient cnn Khalid CNN Efficiency of Other image


    model to detect m.Hosny model in terms forgeries are not
    copy-move image (2022) of time and detected
    forgery accuracy
    2. Image Forgery Ali et al Image comparison of Accuracy
    Detection Using (20 22) recompressing different percentage is less
    Deep learning by algo,Deep techniques. The
    Recompressing learning output was
    Images visually for
    better
    understanding.

    3. Image forgery Hashmi et al CNN High High


    detection using CNN (2023) accuracy,the computation,large
    features are memory footprint
    8 SREYA T(TKR20CS074)
    extracted directly
    10/17/2023
    WHAT IS NEURAL NETWORK ?

    A neural network is a
    machine learning model
    designed to mimic the
    function and structure of
    the human brain.

    SREYA T(TKR20CS074) 10/17/2023


    9
    Class A

    Class B

    Output layer

    Input layer

    Hidden layer

    10 SREYA T(TKR20CS074) 10/17/2023


    TYPES OF NEURAL NETWORK

    SREYA T(TKR20CS074) 11 10/17/2023


    CONVOLUTIONAL NEURAL NETWORK

    •A Convolutional neural network (CNN or convnet) is a subset of


    machine learning. It is one of the various types of artificial
    neural networks that is specifically used for image recognition and
    tasks that involve the processing of pixel data.

    •Convolutional Neural Networks (ConvNets or CNNs) are more often


    utilized for classification and computer vision tasks.

    12 SREYA T(TKR20CS074) 10/17/2023


    TYPES OF CNN

    Convolutional layer

    Pooling layer

    Fully-connected (FC) layer

    13 SREYA T(TKR20CS074) 10/17/2023


    PROPOSED METHOD
     SYSTEM ARCHITECTURE

    login/register
    display the output to input
    the user image(jpeg,pngetc) Front
    End

    Test case Back


    End

    Pre-processing
    Feature extraction

    CNN Algorithm
    SREYA T(TKR20CS074) 10/17/2023 14

    Result
    EXPLANATION

     The user uploads an image on the website front-end.

     The image is evaluated by the CNN model, which has been


    developed by training and testing the dataset.

     The dataset goes through a pre-processing stage where any


    noisy data present in the dataset is cleaned and null values
    are removed.

    SREYA T(TKR20CS074) 10/17/2023


    15
     After the pre-processing stage, the image is passed on to the
    feature extraction stage.

     In the feature extraction stage, the required features are kept and
    the other unimportant features are removed for further stages.

     The extracted features are then passed on to the classification


    stage, where the CNN model classifies the image as either
    authentic or manipulated.

    SREYA T(TKR20CS074) 10/17/2023


    16
    DATASET DETAILS

     The dataset used in the system consists of two folders –


     Authenticated images.
     Tampered images.

     The number of authenticated images in the dataset is around


    7000, while the number of tampered images is around 3000.

    10/17/2023
    17 SREYA T(TKR20CS074)
     The model must be trained with data that contains
    different types of image forgery.

     The custom dataset was created to meet this requirement


    and contains images with these types of tampering.

     Two datasets – CASIA 2.0 and MICC F200 [14] is


    combined and this combination is used to train the
    model.

    SREYA T(TKR20CS074) 18 10/17/2023


    WORKING WITH CNN

     CNN algorithm is used to classify an image as either authentic or


    forged based on the features extracted from the preprocessed
    images.

     The features are learned through a process of convolution, pooling,


    and activation functions.
    CONVOLUTIONAL LAYER
    The convolution layer applies a set of learnable filters to the input
    image to extract relevant features

    SREYA T(TKR20CS074) 19 10/17/2023


    1×1+0×2+3×1+4×0+5×1+6×1+7×1+8×0+9×1=31

    10/17/2023
    20 SREYA T(TKR20CS074)
     POOLING
     The pooling layer reduces the spatial dimensions of the feature maps
    and helps to extract the most important info from the features.

     The activation function applies a nonlinear transformation to the


    output of the convolution and pooling layers.

     After feature extraction, the output of the network is passed through a


    fully connected layer

     The softmax activation function is used.


    SREYA T(TKR20CS074) 10/17/2023
    21
    ERROR LEVEL ANALYSIS (ELA)

     Error Level Analysis (ELA) is a digital image analysis technique


    used to identify potential areas of manipulation in an image.
     ELA works by analyzing the differences in the error levels between
    different regions of an image.
     Error levels refer to the amount of compression artifacts present in
    an image after it has been compressed and saved in a lossy format
    such as JPEG.

    SREYA T(TKR20CS074)
    10/17/2023
    22
    ALGORITHM OF ELA
     We take the forged image A (images shown in Figure 1b
    tamper images)
     Then recompress it, let us call the recompressed image
    as 𝐴𝑟𝑒𝑐𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑒(images shown in Figure 1c are
    recompressed forged images).
     Now we take the difference of the original image and the
    recompressed image, let us call it 𝐴𝑑𝑖𝑓𝑓(images shown in
    Figure 1e are the difference of Figure 1b,c, respectively).
     Now due to the difference in the source of the forged part
    and the original part of the image, the forged part gets
    highlighted in 𝐴𝑑𝑖𝑓𝑓 (as we can observe in Figure 1d,e,
    respectively).
    SREYA T(TKR20CS074) 23 10/17/2023
    • We train a CNN-based
    network to categorize an
    image as a forged image
    or a genuine one
    using 𝐴𝑑𝑖𝑓𝑓 as our input
    features.

    • Figure gives the


    pictorial view of the
    overall working of the
    proposed method.

    10/17/2023
    24 SREYA T(TKR20CS074)
    RESULT

     Accuracy of model
     The system achieved a
    high detection
    accuracy of 93%.

    10/17/2023
    25 SREYA T(TKR20CS074)
    CALCULATION

    Following terms are initially calculated for the evaluation:


     Total_Images: The total number of images that were
    tested.
     TP (true positive): Correctly identified tampered images.
     TN (true negative): Correctly identified genuine images.
     FN (false negative): Wrongly identified tampered images,
    the tampered images which have been identified as
    genuine images.
     FP (false positive): Wrongly identified genuine images,
    the genuine images which have been identified as
    tampered images.

    26 10/17/2023
    SREYA T(TKR20CS074)
    shows the training validation loss and training validation
    accuracy of our model.

    27 10/17/2023
    SREYA T(TKR20CS074)
    shows the confusion matrix achieved while training our model

    28 10/17/2023
    SREYA T(TKR20CS074)
    SREYA T(TKR20CS074) 29 10/17/2023
    FUTURE SCOPE

     Combination with Other Techniques


     Handling Spoofing: Future research could enhance the proposed
    technique to handle other types of image forgeries, such as spoofing.
     Handling Tiny Images: Enhance the proposed technique to work well
    for tiny images, which would be particularly useful in scenarios where
    low-resolution images are encountered.
     Development of Extensive Image Forgery Database

    10/17/2023
    30
    SREYA T(TKR20CS074)
    CONCLUSION

     The CNN model was trained and tested with a dataset of


    approximately 9500 images and accuracy of 93% is obtained.
     Its accuracy, speed, and versatility make it suitable for various
    applications.
     Its potential for enhancing the integrity and credibility of digital
    images cannot be overstated.
     The system outperformed existing methods for image forgery
    detection.

    10/17/2023
    31 SREYA T(TKR20CS074)
    REFERENCES

    [1] M. Patel, K. Rane, N. Jain, P. Mhatre and S. Jaswal, "Image Forgery Detection using
    CNN," 2023 3rd International Conference on Intelligent Technologies (CONIT), Hubli, India,
    2023, pp. 1-4, doi: 10.1109/CONIT59222.2023.10205377.
    [2] T. Singh, Y. Goel, T. Yadav and S. Seniaray, "Performance Analysis of ELA-CNN model for
    Image Forgery Detection," 2023 4th International Conference for Emerging Technology
    (INCET), Belgaum, India, 2023, pp. 1-6, doi: 10.1109/INCET57972.2023.10170007.
    [3] Ranjan, Shruti, Prayati Garhwal, Anupama Bhan, Monika Arora, and Anu Mehra. ”Framework
    for image forgery detection and classification using machine learning.” In 2018 2nd
    International Conference on Trends in Electronics and Informatics (ICOEI), pp. 1- 9. IEEE,
    2018.
    [4] Ali, S. S., I. I. Ganapathi, N. S. Vu, S. D. Ali, N. Saxena, and N. Werghi. ”Image Forgery
    Detection Using Deep learning by Recompressing Images. Electronics 2022, 11, 403.” (2022).

    32 SREYA T(TKR20CS074) 10/17/2023


    THANK YOU
    33 10/17/2023
    SREYA T(TKR20CS074)
    SREYA T(TKR20CS074) 34 10/17/2023

    You might also like