DIGITAL IMAGE

FORGERY DETECTION
ABSTRACT:

With the rise of digital images on social media, detecting forged images is crucial to prevent the spread
of fake information. Traditional forgery detection techniques focus on detecting only one type of
forgery (e.g., image splicing or copy-move), which is not practical for real-world applications. Elements
from the same image are copied and pasted within the image itself to hide or duplicate certain parts.
Parts of an image are taken from a different image and merged to create a manipulated
composite .Copy-move involves duplication within the same image, while splicing introduces elements
from an external source, often leading to inconsistencies in lighting and texture.
INRODUCTION:
• Image forgery techniques are mainly categorized into copy-move and splicing.

• Advanced editing techniques make it challenging to verify the authenticity and reliability of digital
images.
• Forged images can appear visually realistic, making them hard to distinguish from real ones.

• Image forensics help verify authenticity, ownership, and copyright while detecting alterations.

• A pre-trained model is fine-tuned using feature-enhanced images from original and compressed
versions.
• Among eight tested models, MobileNetV2 achieves the highest accuracy (~95%) with faster
training.
• Realistic visual effect can be achieved in such a way that forged images are visual
indistinguishable from real ones.
TYPES OF IMAGE FORGERY
LITERATURE REVIEW :
Paper Title Authors Methodologies Key findings Drawbacks
:
Literature survey of image
forgery detection using ML.
Sirajum Monira Bipaspha Reviews ml-based forgery
detction, incluing deep
Highlights the role of Ai-driven
forgery and need to
Generalizatin issues and
adversarial robustness
learning, and blockchain explainabl AI models. remain challenges.
authentication.

Digital image forgery Shwetha B, Covers active and passve Identifies strengths and Limited discussion on
Techiques: S.V.Sathyanarayana authentication models , weakness of various forgery deeplearning based
including copy-move and techniques . approaches.
splicing techiques

Detection of copy-move Fidrich, J; A method to copy-move Detects forgery even when May struggle with highly
forgery in digital images Soukal, D; forgery by analyzing copied region is retouched complex forgeries or with
Lukas, J; duplicated regions and their and saved in lossy formats. heavy post-processing.
modifications.

Digital Image detection based Yerushalmy exploits artifacts introduced Does not require additional Sensitive to natural variations
on lens and sensor by optical and sensing data, such as watermarks, in camera characteristics,
Abberration systems, such as chromatic , and does not rely on training which may affect detection
to authenticate images. with other images. accuracy.

Asurvey on Digital Image Deependra kumar shukla, Reviews Blind- forgery Discusses various image Challenges in dataset
forensic methods based on Abhishek Bansal, Pawan detection techinques, forgery methos and their availability and real-world
Blind Forgery detection Singh. including splicing, forensic applications. applications.
retouching,morphing.
PROPOSED SYSTEM :

• The rise of digital image manipulation tools has led to increased image forgeries, making
detection methods essential.
• This project employs Python and a CNN-based model, achieving 98% training accuracy and
92% validation accuracy in distinguishing real and tampered images.
• A dataset of 12,615 images (7,492 real, 5,123 fake) provides a comprehensive evaluation set
for reliable performance testing
PROPOSED SYSTEM :

• Error Level Analysis (ELA) is used as a preprocessing step, resizing images to 256×256 and
detecting inconsistencies in compression levels that indicate tampering.
• Processed images are converted into numpy arrays, allowing the CNN model to leverage
both deep learning techniques and ELA insights for enhanced detection accuracy.
• By combining CNNs and ELA, the project significantly improves digital image forgery
detection, with potential applications in forensics, media verification, and cybersecurity.
PROPOSED COPY-MOVE
FORGERY DETECTION :
 SYSTEM REQUIREMENTS:
• HARDWARE REQUIREMENTS:
• System : Pentium i3 Processor.

• Hard Disk : 500 GB.

• Monitor : 15’’ LED.

• Input Devices : Keyboard, Mouse.

• Ram : 8 GB.
SOFTWARE REQUIREMENTS

• Operating system : Windows 10 / 11.

• Coding Language : Python 3.10.9.

• Web Framework : Flask.

• Frontend : HTML, CSS, JavaScript.

we KAZE features, a novel method for multiscale 2D feature detection and description in nonlinear scale
spaces. In contrast to previous ap_x0002_proaches that rely on the Gaussian scale space, our method is
based on nonlinear scalespaces using efficient AOS techniques and variable conductance diffusion. Despite
ofmoderate increase in computational cost, our results reveal a step forward in perfor_x0002_mance both in
detection and description against previous state-of-the-art methods suchas SURF, SIFT or CenSurE.
In the next future we are interested in going deeper i n nonlinear diffusion filteringand its applications for
feature detection and description. In particular, we think thathigher quality nonlinear diffusion filtering such as
coherence-enhancing diffusion fil_x0002_tering [21] can improve our current approach substantially. In
addition, we will workin the direction of speeding-up the method by simplifying the nonlinear diffusion
pro_x0002_cess and by using GPGPU programming for real-time performance