XXX-X-XXXX-XXXX-X/XX/$XX.

00 ©20XX IEEE
 A Systematic Review of Fake News Detection
Using Machine Learning and Deep Learning
Aniket Sharma Yuvraj Pandey Pranay Bhurewar
Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Lovely Professional University Lovely Professional University Lovely Professional University
Phagwara, Punjab, India Phagwara, Punjab, India Phagwara, Punjab, India
[email protected] [email protected] [email protected]
Priyansh Tiwari Shashwat Shriwas Patchipulusu Leeladharnath
Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Lovely Professional University Lovely Professional University Lovely Professional University
Phagwara, Punjab, India Phagwara, Punjab, India Phagwara, Punjab, India
[email protected] [email protected] [email protected]

information sharing. However, this has also fueled the rise of

I. ABSTRACT-- THIS RAPID ACCELERATION IN INFORMATION fake news—deliberately false information designed to
SHARING, ON THE ONE HAND, HAS FUELLED FAKE NEWS, mislead readers. Social media amplifies its spread, often
WHICH HAS UNDERMINED TRUST AND THREATENS outpacing fact engagement with credibility insights, and
DEMOCRATIC PROCESSES. FAKE NEWS DETECTION HAS BEEN scalability for large datasets and evolving misinformation
ACKNOWLEDGED AS ONE OF THE CRITICAL AREAS OF trends.
RESEARCH BECAUSE IT INVOLVES ARTIFICIAL INTELLIGENCE
TECHNIQUES SUCH AS NLP, ML, AND SOCIAL NETWORK III. PROJECT DESCRIPTION
ANALYSIS TO COMBAT MISINFORMATION. IT TALKS ABOUT
THE EFFICACY OF ALGORITHMS, FEATURE SELECTION, AND
We are developing a Fake News Detection System to
combat the spread of misinformation in the digital age. The
USER BEHAVIOUR IN SPREADING RUMOURS AND BRINGS OUT
ETHICAL CONSIDERATIONS LIKE TRANSPARENCY AND
system will use natural language processing (NLP) and
ACCOUNTABILITY. IT UNDERLINES INTERDISCIPLINARY
machine learning techniques to analyze news articles,
COOPERATION TO STRENGTHEN DETECTION SYSTEMS AND
social media posts, and other digital content. By identifying
DEVELOP A HEALTHIER INFORMATION ECOSYSTEM TO
intentionally misleading or verifiably false information, the
CHECKING EFFORTS. THIS LEADS TO MANIPULATION OF
project aims to provide users with a reliable tool to verify
news credibility. The system will feature a user-friendly
PUBLIC OPINION, EROSION OF TRUST IN LEGITIMATE MEDIA,
AND DISTORTION OF DEMOCRATIC PROCESSES. THE SOCIETAL
interface and integrate real-time data from multiple sources
IMPACT OF FAKE NEWS IS PROFOUND, PARTICULARLY DURING
to ensure accuracy and timeliness. Project Scope
The project focuses on enhancing fake news detection
ELECTIONS OR PUBLIC HEALTH CRISES. IT CAUSES
through advanced algorithms like CNNs and RNNs, feature
CONFUSION, FEAR, AND POLARIZATION WHILE CONTRIBUTING
extraction using linguistic patterns and metadata, real-time
TO "INFORMATION OVERLOAD," MAKING CREDIBLE SOURCES
analysis of social media content, use supervised machine
HARDER TO IDENTIFY. TO COMBAT THIS, RESEARCHERS ARE learning for fake news detection, highlighting linear
LEVERAGING MACHINE LEARNING (ML) AND NATURAL classification's effectiveness and the need for advanced
LANGUAGE PROCESSING (NLP) TO DEVELOP ALGORITHMS algorithms with big data.[8] The paper proposes a Bi-
FOR DETECTING MISINFORMATION. THESE SYSTEMS ANALYZE directional LSTM model with GloVe embeddings for fake
TEXTUAL FEATURES, ENGAGEMENT PATTERNS, AND news detection, outperforming other methods and
CONTEXTUAL CUES TO ASSESS CREDIBILITY. FOR INSTANCE, emphasizing adaptive learning rates for improved
ML MODELS TRAINED ON LARGE DATASETS CAN IDENTIFY accuracy.[9] This project applies machine learning
PATTERNS INDICATIVE OF FALSE CONTENT, WHILE NLP techniques like Random Forest, Logistic Regression, and
EVALUATES SENTIMENT AND STRUCTURAL ANOMALIES IN Decision Tree to detect fake news on social media platforms
ARTICLES. DESPITE ADVANCEMENTS, CHALLENGES PERSIST like YouTube.[10] The paper proposes a fake news detection
AS MISINFORMATION TACTICS EVOLVE, COMPLICATING method using Web Scraping and Text Similarity, with
DETECTION EFFORTS. ETHICAL CONCERNS ABOUT potential improvements through machine learning on
CENSORSHIP, ALGORITHMIC BIAS, AND IMPACTS ON FREE Android and iOS platforms.[11] The paper proposes a fake
SPEECH FURTHER COMPLICATE IMPLEMENTATION. news detection based on machine learning system with a
BALANCING EFFECTIVE MISINFORMATION MITIGATION WITH trust and reputation framework, comparing traditional
THE PROTECTION OF DEMOCRATIC VALUES REMAINS
databases to Blockchain solutions for news reliability.[12]
The paper proposes a framework using XGBoost and LSTM-
CRUCIAL.
ATT models for detecting fake news, with future research
exploring social media content and spatiotemporal data.[13]
The paper uses machine learning classifiers, with SVM
achieving 93.61% accuracy in fake news detection,
II. INTRODUCTION highlighting the need for automated algorithms while noting
The digital era has revolutionized news consumption, limitations in feature selection and deep learning models.[14]
with the internet and social media enabling instant The paper explores fake news detection using the 2016 U.S.
Elections dataset, highlighting the importance of text machine learning model for fake news detection on Twitter,
preprocessing and EDA, with transformer-based models outperforming SVM and Naïve Bayes, and suggests
outperforming others in accuracy and computation time.[15] incorporating domain knowledge for future
The study uses NLP techniques with TensorFlow and Keras improvements.[6] The paper uses AI and ML for binary
to detect fake news on social media, achieving high classification of social media news, helping users identify
performance with news content and titles, and suggesting fake news and verify website legitimacy with the most
future improvements with LSTM and international accurate model.[7] The paper proposes model using NLP,
implementation.[16] The project proposes a fake news with the SVM model achieving 98% accuracy, and plans to
identification system using deep learning and NLP,
incorporate deep learning and new datasets in future
achieving 90% accuracy, with future improvements
work.[19] The study proposes an approach via machine
suggested through larger datasets.[17] The paper discusses
the challenges of distinguishing false information on social learning for the detection of fake news, outperforming
media and highlights the use of deep learning, beyond existing systems by evaluating both content and
machine learning, in applications like audio and speech aesthetics.[20] The paper proposes using the Naive Bayes
processing for fake news detection.[18] The study proposes classification model for detecting fake news on Facebook,
an ML-based fake news detection model using NLP, with the improving accuracy and AUC scores, with potential for
SVM model achieving 98% accuracy, and plans to further enhancement using machine learning techniques.[21]
incorporate deep learning and new datasets in future proposed techniques and the SVM algorithm. They also
work.[19] The study proposes an approach via machine employed NLP algorithm to process the news in the
learning for the detection of fake news, outperforming segments that can be classified using keywords and can be
existing systems by evaluating both content and classified as fake or real. TABLE 1 includes references to
aesthetics.[20] The paper proposes using the Naive Bayes the relevant work and discoveries made by different authors.
classification model for detecting fake news on Facebook,
improving accuracy and AUC scores, with potential for
further enhancement using machine learning techniques.[21] V. RESEARCH QUESTIONS
proposed techniques and the SVM algorithm. They also
employed NLP algorithm to process the news in the
segments that can be classified using keywords and can be
Q1) How can model performance be improved on
classified as fake or real. TABLE 1 includes references to the less-balanced datasets in fake detection?
relevant work and discoveries made by different authors.
To tackle data imbalance, techniques such as
IV. REVIEW OF LITERATURE Synthetic Minority Over-sampling Technique
With major ramifications for public opinion, (SMOTE) or class-weight adjustments can be
social cohesion, and democratic institutions, the integrated to minimize bias toward majority
spread of false information and fake news on classes. This ensures that models are equally
digital platforms has emerged as a major sensitive to real and fake news categories,
worldwide concern. It is getting more and more potentially enhancing overall detection accuracy.
difficult to tell real news from fake as information Q2) How can explainability in fake news
becomes more widely available. The investigation detection models be enhanced to improve users
of machine learning and deep learning techniques trust and accuracy?
to automate the detection of fake news has been
spurred by this growing concern. This study uses an Enhancing explainability using attention
LSTM neural network with GloVe embeddings, mechanisms or model interpretation methods like
tokenization, and N-grams for fake news detection,
SHAP values would help users understand why
achieving 99.88% accuracy and outperforming other
methods, with accuracy improving as epochs increase. [1] content is flagged as fake, building trust and
This study leverages NLP, machine learning, and deep enabling users or moderators to take appropriate
learning techniques, including CNN, LSTM, Naïve Bayes, actions with greater confidence.
and Decision Tree Regressor, for fake news detection,
showcasing improved accuracy and providing a Q3) How can fake news detection models be
comprehensive review of existing methods.[2] This study improved to generalized better across different
uses TF-IDF bi-grams and probabilistic context-free
grammar (PCFG) on 11,000 articles on how to identidy and
news domains?
detect fake news, achieving 77.2% accuracy in identifying
non-credible sources, with PCFG slightly improving Current models often excel within specific
recall.[3] Machine learning algorithms, particularly Naïve domains (e.g., politics) but struggle with cross-
Bayes and Decision Tree, excel in fake news detection, with domain generalization. By integrating transfer
Naïve Bayes achieving the highest F1-Score of 99.72%.[4] learning and domain adaptation techniques,
The paper proposes using Naïve Bayes classification for models can retain accuracy and handle varying
fake news detection on social media, emphasizing dataset
updates to improve accuracy.[5] The study proposes a
language styles across domains, improving overall Optimization and improvement of fake news
adaptability and reliability. detection using deep learning approaches for
societal benefit,” Int. J. Inf. Manag. Data Insights,
Q4) How can the feature extraction process be vol. 1, no. 2, p. 100051, 2021, doi:
10.1016/j.jjimei.2021.100051.
enhanced to capture more complex linguistic and
semantic patterns in fake news? [2] R. K. Kaliyar, “Fake News Detection Using A Deep
Neural Network,” 2018 4th Int. Conf. Comput.
Current feature extraction relies on methods like Commun. Autom., pp. 1–7, 2018.
TF-IDF and n-grams, which may miss nuanced
[3] L. Dyson, A. Golab, and S. Gilda, “Evaluating
linguistic cues. By incorporating advanced NLP Machine Learning Algorithms for Fake News
techniques like sentiment analysis, stance Detection,” 2017 IEEE 15th Student Conf. Res.
detection, and rhetorical structure analysis, Dev., no. December, pp. 110–115, 2017, doi:
models could gain a deeper understanding of the 10.1109/SCORED.2017.8305411.
content, improving fake news detection accuracy. [4] V. Gupta, R. S. Mathur, T. Bansal, and A. Goyal,
"Fake News Detection using Machine Learning,"
2022 Int. Conf. Mach. Learn. Big Data, Cloud
Parallel Comput., vol. 1, no. May, pp. 84–89, 2022,
Conclusion and Future Scope doi: 10.1109/COM-IT-CON54601.2022.9850560.
This study uses the Naive Bayes classification [5] A. Jain and A. Kasbe, “Fake News Detection,” 2018
model to demonstrate an efficient method for IEEE Int. Students’ Conf. Electr. Electron. Comput.
identifying false information on social networking Sci., pp. 1–5, 2018.
sites. The improved AUC values show that the
[6] E. M. Mahir, S. Akhter, and M. R. Huq, “Detecting
suggested approach greatly increases the accuracy Fake News using Machine Learning and Deep
of differentiating between genuine and fraudulent Learning Algorithms,” 2019 7th Int. Conf. Smart
posts. This study demonstrates the possibility of Comput. Commun., pp. 1–5, 2019.
automating the detection of false information in
[7] V. Kumar, A. Kumar, A. K. Singh, and A. Pachauri,
social media content by utilizing machine learning “Fake News Detection using Machine Learning and
algorithms. Natural Language Processing,” 2021 Int. Conf.
Technol. Adv. Innov., pp. 547–552, 2021, doi:
Ref Datasets
Accura
10.1109/ICTAI53825.2021.9673378.
Classification Type Models used Charactersti
cy(%)
cs [8] B. Al Asaad, “A Tool for Fake News Detection,”
Multiclass Kaggle
[3] Bi-LSTM Dataset 91.50 2018 20th Int. Symp. Symb. Numer. Algorithms Sci.
Classification
Comput., pp. 379–386, 2018, doi:
Binary Fake News
[5] RoBERTa Corpus 96.20 10.1109/SYNASC.2018.00064.
Classification
Essemble FakeNews
Multiclass [9] P. Bahad, P. Saxena, and R. Kamal, “ScienceDirect
[6] of CNN Net Dataset 92.80
Classification ScienceDirect Fake News Detection using Bi-
and GRU
Binary LIAR directional LSTM-Recurrent Neural Fake News
[7] DistilBERT Dataset 94.30 Detection using Bi-directional LSTM-Recurrent
Classification
Binary Transforme GossipCOP Neural Network Network,” Procedia Comput. Sci.,
[8] Dataset 93.70
Classification r-XL vol. 165, no. 2019, pp. 74–82, 2020, doi:
Multiclass Capsule Kaggle 10.1016/j.procs.2020.01.072.
[11] Dataset 90.00
Classification Network
Binary Polyfact [10] K. Shanmugavadivel, M. Subramanian, R. Sanjai,
[13] XLNet 95.10
Classification Dataset M. B. Sameer, and K. Motheeswaran, “Beyond
Multiclass
Hierarchial Kaggle Tech@DravidianLangTech2024: Fake News
[15] Attention Dataset 89.50 Detection in Dravidian Languages Using Machine
Classification
Net
Learning,” DravidianLangTech 2024 - 4th Work.
Fake News
Binary Speech, Vision, Lang. Technol. Dravidian Lang.
[20] ALBERT Challenge 94.00
Classification
Dataset Proc. Work., pp. 124–128, 2024.

[11] S. Lohitha, S. D. Reddy, B. R. Krishna, and N. S.

REFERENCES Krishna, “Fake News Detection Using Machine
Learning,” Lect. Notes Networks Syst., vol. 645
[1] T. Chauhan, M. E, and H. Palivela, “International LNNS, no. April, pp. 463–470, 2023, doi:
Journal of Information Management Data Insights 10.1007/978-981-99-0769-4_41.
[12] C. Marche, I. Cabiddu, C. G. Castangia, L. Serreli, [17] A. Matheven and B. V. D. Kumar, "Fake News
and M. Nitti, “Implementation of a Multi-Approach Detection Using Deep Learning and Natural
Fake News Detector and of a Trust Management Language Processing," 2022 9th Int. Conf. Soft
Model for News Sources,” IEEE Trans. Serv. Comput. Mach. Intell. ISCMI 2022, pp. 11–14,
Comput., vol. 16, no. 6, pp. 4288–4301, 2023, doi: 2022, doi: 10.1109/ISCMI56532.2022.10068440.
10.1109/TSC.2023.3311629.
[18] S. I. Manzoor, J. Singla, and Nikita, "Fake news
[13] J. Lin, G. Tremblay-Taylor, G. Mou, D. You, and K. detection using machine learning approaches: A
Lee, “Detecting Fake News Articles,” Proc. - 2019 systematic review," Proc. Int. Conf. Trends
IEEE Int. Conf. Big Data, Big Data 2019, pp. 3021– Electron. Informatics, ICOEI 2019, no. Icoei, pp.
3025, 2019, doi: 230–234, 2019, doi: 10.1109/ICOEI.2019.8862770.
10.1109/BigData47090.2019.9005980.
[19] M. Aljabri, D. M. Alomari, and M. Aboulnour,
[14] N. Fahad et al., “Stand up Against Bad Intended “Fake News Detection Using Machine Learning
News: An Approach to Detect Fake News using Models,” Proc. - 2022 14th IEEE Int. Conf.
Machine Learning,” Emerg. Sci. J., vol. 7, no. 4, pp. Comput. Intell. Commun. Networks, CICN 2022, pp.
1247–1259, 2023, doi: 10.28991/ESJ-2023-07-04- 473–477, 2022, doi:
015. 10.1109/CICN56167.2022.10008340.

[15] S. Mondal, P. Srinivasan, and M. E. Ahammed, [20] M. A. Shaik, M. Y. Sree, S. S. Vyshnavi, T. Ganesh,
“Fake News Detection: A Comparative Study of D. Sushmitha, and N. Shreya, “Fake News
Machine Learning Techniques,” 2023 Int. Conf. Detection using NLP,” Int. Conf. Innov. Data
Comput. Intell. Networks Secur. ICCINS 2023, pp. Commun. Technol. Appl. ICIDCA 2023 - Proc., pp.
1–6, 2023, doi: 399–405, 2023, doi:
10.1109/ICCINS58907.2023.10450148. 10.1109/ICIDCA56705.2023.10101305

[16] S. Sharma, M. Saraswat, and A. K. Dubey, “Fake [21] A. Jain and A. Kasbe, “Fake News Detection,” 2018
News Detection Using Deep Learning,” Commun. IEEE Int. Students’ Conf. Electr. Electron. Comput.
Comput. Inf. Sci., vol. 1459 CCIS, pp. 249–259, Sci. SCEECS 2018, pp. 1–5, 2018, doi:
2021, doi: 10.1007/978-3-030-91305-2_19. 10.1109/SCEECS.2018.8546944.