By: C S Nakul Kalyan, Asia University, Taiwan

Abstract

Due to the advancement of deepfake technologies, the problem of deepfake social media profiles and identity theft, which is done by Artificial Intelligence, has increased. A multilayered detection framework has been implemented due to the misuse of these advancements. To detect the fake profiles, Machine learning models such as Decision trees, Random forests, and Support vector Machines (SVMs) are used to detect the activity patterns, network features, and the information of the profiles. Deep learning based detection techniques, such as CNNs and GAN-Based Detectors, have the potential to detect deepfake images and videos at the content level with high accuracy. The integration of Blockchain makes use of smart contracts and localized validation to give authenticity and transparency at the system level. These strategies, when they are combined, give a flexible way to reduce identity theft and increase trust in the public in using the online platforms.

Keywords

Fake Media Profiles, Identity Theft, Deepfake Detection, Artificial Intelligence (AI), Blockchain Technology.

Introduction

The Growth of Online Social Networks (OSNs) has changed digital communications, which allows for unparalleled connectivity and sharing of information with connections. This act has made it easier for the fraudsters to create fake pro- files, spread misinformation, and conduct identity theft. These advancements in Artificial Intelligence have increased the threats by letting the production of realistic synthetic fake identities and manipulated media content. These practices take out the individual’s privacy and reduce the trust and security to using the online platforms. The main goal of this study is to create an automated

detection system that can adapt to evolving attack techniques. The ML Models, such as the Support Vector Machines (SVMs), Decision Trees, and Random Forests, have been used to detect fake accounts by checking the metadata, net- work features, and user activities [1][2]. With these deep learning techniques, such as Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs)-modal detectors have been showing a good accuracy in detecting deepfake images and videos by verifying the digital artifacts and the irregularities [3][4]. With these, Blockchain technology has been proposed to increase transparency and trust by applying techniques such as decentralized validation, immutable records, and contracts to verify the content, whether it is an original or fake. These approaches together forms a multi-layered frame- work to detect false social media profiles and reduce identity theft in the online platforms. In this article, we will go through these approaches, which give us a method to protect our digital identities and ensure content protection, and will raise trust in using online platforms.

Proposed Methodology

Detecting fake profiles and preventing identity theft requires a multi-layered approach that combines Machine Learning, Deep Learning, and blockchain- enabled verification [1][3]. This study proposes a combined methodology with approaches such as Profile-level detection, content-level deepfake analysis, and system-level verification such as shown in Figure 1.

Profile-Level Detection Using Machine Learning

For detecting the fake profiles, the accounts that are all suspicious have been identified through the process of behavioral and metadata analysis [2]:

Data Collection

The account information, such as user names, bio, Friends, follower count, posting frequency, likes, and comments of the user, has been gathered by using the Social media APIs.

Pre-Processing

After collecting the user information, the process, such as Noise removal, normalization, and feature engineering, has been done, and the data will be pre- processed.

Feature Extraction

The different feature extractions are shown in Figure 2 below,

Profile Features: The profile feature will be extracted from the accounts that contain empty bios and abnormal user names.

Activity Features: The Activity feature will be extracted by the high- posting with little engagement among the followers.

Network Features: The network features have been extracted from the profiles containing skewed friend-to-follower ratios or profiles containing isolated groups.

Classification Modals

The supervised learning algorithms, such as the Decision Trees, Random Forest, and Support Vector Machines (SVMs), were applied to classify the profiles, whether they are genuine or fake.[1][2].

Content Level Deepfake Detection Using Deep Learn- ing

Identity Theft, which is done in the modern platform, frequently uses AI- generated deepfake images or videos [5], and to detect this kind of content, a second detection layer is applied, such as:

Image and Video Processing

The image and video processing is done by extracting frames, detecting the facial regions from the data, normalizing them, and convert the content into a consistent format in which we can apply deep learning models to detect the deepfakes.

Deep Learning Modals

The deep learning models that are applied here are shown below:

Convolutional Neural Networks (CNNs): The CNNs have been used to detect the pixel-level inconsistencies, which include unnatural lightning, texture irregularities, or irregular blending artifacts [3].

Generative Adversarial Network (GAN)-Based Detection: The GAN- based detection model is used for identifying fake content by checking noise and statistical patterns in the data.

XGBoost and Ensemble Modals: These modals are used to enhance the accuracy of detecting deepfakes on mixed datasets in real-life platforms, such as Instagram and Twitter.

Performance

The performance of the above deep learning models has given above 95 percent of accuracy as shown in Table 1.

Table 1: Deep Learning Modal Architecture and Performance

Modal Type	Architecture	I/P Size	Parameters	Accuracy	Processing Time(ms)
CNN (basic)	3 conv+2 FC Layers	224x224x3	2.3M	91.4	15
CNN (ResNet-50)	50-layer residual network	224x224x3	25.6M	95.8	35
GAN Detector	Discriminator Network	256x256x3	8.7M	93.2	28
ensemble Modal	CNN+GAN+XGBoost	Variable	15.2M	97.3	45

for differentiating the original and fake profile images and videos, and it provides a great resilience against deep-fake content.

System Level Verification Using Blockchain and Col- lective Validation

A blockchain Validation system has been implemented to ensure long-term trust and prevent the manipulation of content [4]. The system implemented is as follows:

Blockchain Ledger

Blockchain Ledger is being used to save all the details of the profile, the detection results, and the security check that is made on the contents, which will be stored securely.

Smart Contracts

The smart contracts will be used to validate the source of the digital content, and they will provide the security criteria before the content is published or the account verification process.

Collective voting Mechanism

In this Mechanism, multiple AI nodes will be used to assess the identified content and will be used to detect the inconsistencies using methods such as RANDAO, which will eliminate the single-point failures.

Reputation Tracking

Reputation Tracking specifies that nodes can earn trust or lose trust score, which will be totally dependent on their previous accuracy which will be ensuring consistency in the detection process.

Identity Theft and Spoofing Defense

To find a way against identity theft, the following methods have been implemented, such as [3]:

Face Anti-spoofing:

The Face Anti-spoofing is a micro-texture-based deep learning model that will detect replay and mask attacks during the process of validating the profiles.

IoT and Clone ID Security:

This framework is an unsupervised dense Neural network (DNN) that is used to detect inconsistencies on connected platforms, where it will identify and prevents the large-scale digital identity theft.

Conclusion

Deepfake-driven social media content has become a major cause of identity theft by finding the weakness of the authentication process in standard authentication and moderation systems. The study shows that ML models such as Decision Trees, Random Forest, and Support Vector Machines (SVM) can able to detect fake profiles by using the metadata, activity patterns, and their network behavior. On the other side, Deep learning models such as CNNs, GAN-based detectors, can be able to detect manipulated media, in which they can accurately differentiate real and fake information. Along with these techniques, the implementation of blockchain provides a validation mechanism that ensures the verification, content provenance, and transparency across all the online digital platforms. The method that has been used here provides a multi-layered detection framework that will be used to provide protection against online fraud and identity theft. This framework will be more scalable and adaptable to the emerging threats in which it handles the detecting against fake profiles, identity theft, and deepfake manipulations. The future research on this topic can be on reducing false positives and providing a continuous integration across all social digital platforms, which will be useful for digital identity protection in the advancement of AI.

References

1. P Chinnasamy, Ramesh Kumar Ayyasamy, Suthashini Subramaniam, An- buselvan Sangodiah, Aqsa Iftikhar, and Ajmeera Kiran. Fake social media profile identification and report using machine learning. In 2024 Interna- tional Conference on Signal Processing, Computation, Electronics, Power and Telecommunication (IConSCEPT), pages 1–5. IEEE, 2024.
2. AKM Rubaiyat Reza Habib, Edidiong Elijah Akpan, Bhaskar Ghosh, and Indira Kalyan Dutta. Techniques to detect fake profiles on social media using the new age algorithms-a survey. In 2024 IEEE 14th Annual Computing and Communication Workshop and Conference (CCWC), pages 0329–0335. IEEE, 2024.
3. Meghana Lokhande, Prajot Raut, Kiran Gawali, Mrudul Ahirrao, and Ab- hishek Bhande. Artificial intelligence for detecting cyber attacks in deepfake & identity theft. In 2024 8th International Conference on Computing, Com- munication, Control and Automation (ICCUBEA), pages 1–6. IEEE, 2024.
4. M Swarna Sudha, S Manjula, Indira Bharathi, Valarmathi Krishnasamy, and K Vijayalakshmi. Deepfakeguard: Safeguarding digital platforms against fake profiles using ai. In 2025 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL), pages 1293–1299. IEEE, 2025.
5. Kai-Cheng Yang, Danishjeet Singh, and Filippo Menczer. Characteristics and prevalence of fake social media profiles with ai-generated faces. arXiv preprint arXiv:2401.02627, 2024.
6. Gupta, B. B., Misra, M., & Joshi, R. C. (2012). An ISP level solution to combat DDoS attacks using combined statistical based approach. arXiv preprint arXiv:1203.2400.
7. Jain, A. K., & Gupta, B. B. (2018). Rule-based framework for detection of smishing messages in mobile environment. Procedia Computer Science, 125, 617-623.
8. Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET. Journal of Information Security Vol. 4  No. 3 (2013) , Article ID: 34631 , 15 pages

Cite As

Kalyan C S N (2025) Deepfake Social Media Profiles: Identity Theft in the Age of AI, Insights2Techinfo, pp.1

888300cookie-checkDeepfake Social Media Profiles: Identity Theft in the Age of AIyes