Fake News Detection Projects | NLP · BERT · Fact-Check · Misinformation

Fake News Detection Projects for Final Year Students (2026)

Fake news detection combines text classification with credibility signals, stance toward claims, and careful evaluation on imbalanced or politically sensitive data. Student projects range from classical TF-IDF pipelines to BERT fine-tuning and multi-task claim verification.

This page lists 90+ high-impact topics. Tools include scikit-learn, NLTK, spaCy, TensorFlow/Keras, PyTorch, Hugging Face Transformers and datasets such as LIAR, ISOT, FakeNewsNet and Kaggle fake news sets. Ideal for BE, BTech, MTech CS and AI students in Bangalore and across India.

Core Frameworks & Tools

Libraries and models commonly used in academic fake news detection projects.

scikit-learn NLTK / spaCy BERT / Transformers TensorFlow / Keras PyTorch LIAR · ISOT · Kaggle

Best Fake News Detection Topics & Tools (90+)

Grouped by theme. Each topic lists primary tools and typical datasets.

#	Project Topic	Tools · Datasets
📰 Binary & Multi-Class Classification
1	BinFake vs Real News Binary Classifier	TF-IDF + NB/SVM, ISOT
2	BinKaggle Fake News Competition Pipeline	Kaggle Fake News
3	BinMulti-Class: True / Mostly True / False / Pants	LIAR dataset
4	BinThree-Way: Real / Fake / Satire	Custom / public mix
5	BinCOVID-19 Misinformation Classification	COVID fake news sets
6	BinPolitical News Fake/Real Detection	Political subsets
7	BinHeadline-Only vs Full-Text Classification	Ablation study
8	BinShort Social Media Post Fake Detection	Tweet-style text
9	BinCompare Domain: Politics vs Health vs Tech	Domain transfer
10	BinReal-Time Text Input Fake Score Demo	Gradio / Streamlit
📊 Stance Detection
11	StanceAgree / Disagree / Discuss / Unrelated	FNC-1 style labels
12	StanceClaim–Headline Stance Classification	Pair encoding
13	StanceBody–Headline Compatibility Score	Similarity features
14	StanceMulti-Target Stance toward Entities	Target-aware labels
15	StanceStance + Fake Joint Multi-Task Model	Shared encoder
16	StanceCompare Independent vs Joint Stance Models	Ablation table
✅ Claim Verification · Fact-Check
17	ClaimClaim Extraction from News Articles	Sentence ranking
18	ClaimEvidence Retrieval for a Claim	BM25 / embeddings
19	ClaimClaim–Evidence NLI (entail / contradict)	NLI models
20	ClaimEnd-to-End Claim Verification Pipeline	Retrieve + verify
21	ClaimFact-Check Verdict Classification (True/False)	Fact-check corpora
22	ClaimExplainable Fact-Check with Highlighted Evidence	Attention / spans
23	ClaimMulti-Hop Evidence Aggregation Lite	Multi-document
24	ClaimWikipedia-Based Claim Checking Demo	Wiki dump subset
🔗 Clickbait · Headline Analysis
25	ClickClickbait Headline Binary Detection	Clickbait datasets
26	ClickHeadline–Body Consistency Check	Similarity + classifier
27	ClickSensationalism Score from Lexical Features	Lexicons + ML
28	ClickQuestion / Listicle Pattern Detection	Rule + ML hybrid
29	ClickClickbait vs Informative Headline Study	User study lite
30	ClickSocial Engagement Features + Text	Shares / likes proxy
🧩 Features · Linguistics · Style
31	FeatStylometric Features for Fake News	Readability, POS ratios
32	FeatSentiment and Emotion as Credibility Cues	VADER / emotion
33	FeatNamed Entity Density and Patterns	spaCy NER
34	FeatSource / Author Metadata Features	Publisher signals
35	FeatURL / Domain Reputation Features	Domain lists
36	FeatPropaganda Technique Classification Lite	Propaganda corpora
37	FeatPunctuation and Capitalization Patterns	Surface features
38	FeatFeature Importance Ranking (SHAP / chi2)	Explainability
📐 Classical ML Pipelines
39	MLTF-IDF + Multinomial Naive Bayes	sklearn Pipeline
40	MLTF-IDF + Linear SVM / Logistic Regression	sklearn
41	MLn-gram and Character n-gram Features	char n-grams
42	MLEnsemble: Voting / Stacking of Classifiers	sklearn ensemble
43	MLClass Imbalance Handling for Rare Fake Class	Class weights / SMOTE
44	MLCross-Validation and Calibration Study	GridSearchCV
45	MLError Analysis: False Positives vs Negatives	Confusion deep dive
46	MLLearning Curves and Sample Size Effects	sklearn learning_curve
🤖 Deep Learning · BERT · Transformers
47	BERTBERT Fine-Tuning for Fake News Binary	Hugging Face, ISOT
48	BERTDistilBERT / TinyBERT Efficiency Study	Transformers
49	BERTRoBERTa Multi-Class LIAR Fine-Tune	LIAR + RoBERTa
50	BERTLSTM / BiLSTM Fake News Classifier	Keras / PyTorch
51	BERTCNN Text Model for Fake Detection	1D conv + embeddings
52	BERTCompare Classical vs BERT on Same Split	Accuracy / F1 table
53	BERTHierarchical Attention Network Lite	Word + sentence attn
54	BERTMulti-Task: Fake + Stance Joint Training	Shared BERT head
55	BERTDomain-Adaptive Pretraining on News	Continued MLM
56	BERTAttention Visualization for Interpretability	BertViz concepts
📱 Social · Multimodal · Propagation
57	SocSocial Context Features (shares, comments)	Engagement proxies
58	SocUser Credibility / Bot-like Signals Lite	Account features
59	SocImage + Text Multimodal Fake Detection	CLIP / dual encoder
60	SocPropagation Graph Features Overview	Network stats
61	SocTemporal Patterns: Early Detection Focus	Time-limited features
62	SocCross-Platform Consistency Check	Multi-source text
🏥 Domain-Specific Misinformation
63	DomHealth / Medical Misinformation Classifier	Health rumor sets
64	DomClimate / Environment Misinfo Detection	Domain corpus
65	DomElection-Related Fake News Pipeline	Political sets
66	DomFinancial Scam / Fake News Hybrid	Finance text
67	DomDisaster / Emergency Misinfo Rapid Filter	Event-focused data
68	DomRegional Language Fake News (one language)	Indic / local corpus
⚖️ Evaluation · Ethics · Robustness
69	EvalPrecision–Recall Focus on Fake Class	PR curves
70	EvalTemporal Split vs Random Split Study	Time-aware eval
71	EvalCross-Dataset Generalization Test	Train A, test B
72	EvalBias Audit Across Topics / Sources	Stratified metrics
73	EvalAdversarial Paraphrase Robustness	Paraphrase attacks
74	EvalHuman-in-the-Loop Review Interface	Annotation UI
75	EvalEthics Statement and Limitation Report	Documentation
76	EvalFalse Positive Cost Analysis Case Study	Impact narrative
🔬 Advanced · Deployment · Research
77	AdvActive Learning for Fake Label Efficiency	Uncertainty sampling
78	AdvData Augmentation for Minority Fake Class	EDA / back-translate
79	AdvKnowledge Graph Features for Credibility	Entity linking lite
80	AdvFew-Shot Fake Detection with Prompts	Prompt templates
81	AdvModel Distillation for Mobile Inference	Teacher → student
82	AdvONNX / Quantized Deployment Package	ONNX Runtime
83	AdvStreamlit / Gradio Fake News Checker App	Web UI
84	AdvFastAPI Prediction + Logging Service	REST + metrics
85	AdvLIME / SHAP Explanations for Decisions	lime, shap
86	AdvBenchmark: 3 Datasets × 3 Model Families	Unified eval
87	AdvEarly Detection: Limited Context Setting	Prefix-only text
88	AdvCross-Lingual Fake News Transfer	mBERT / XLM-R
89	AdvEducational Lab: Classical → BERT Path	Curriculum package
90	AdvEnd-to-End: Data → Train → Evaluate → Deploy → Report	Full pipeline
91	AdvReproducibility: Seeds, Configs, Experiment Log	YAML + tracking
92	AdvThesis Package: Methods, Ethics, Results, Discussion	Full documentation

Topics reflect misinformation and fake news research practice with open-source tools and public data. Contact us for pipelines, evaluation metrics, university-format report, PPT and viva Q&A for any topic above.

Why Choose Us for Fake News Detection Projects?

Bangalore-based guidance for BE, BTech and MTech students working on classification, stance and fact-checking systems.

Classification

Binary and multi-class fake/real pipelines with classical and deep models.

Stance Detection

Agree/disagree/discuss labels linking claims to headlines and bodies.

Claim & Fact-Check

Evidence retrieval and NLI-style verification pipelines.

BERT & Robustness

Transformer fine-tuning, explainability and cross-dataset evaluation.

Frequently Asked Questions — Fake News Detection

Top topics include binary fake/real classification, multi-class credibility labels, stance detection (agree/disagree), clickbait detection, claim verification pipelines and BERT fine-tuning on public fake news datasets.

scikit-learn, NLTK/spaCy, TensorFlow/Keras, PyTorch, Hugging Face Transformers (BERT); datasets include LIAR, FakeNewsNet, ISOT, Kaggle Fake News, COVID-19 misinformation sets and similar public corpora.

Yes. Packages include preprocessing pipelines, model training notes, evaluation metrics, university-format report, PPT and viva Q&A.

Fake news tasks often need credibility cues, source features, stance toward claims, and careful handling of political or health misinformation. Evaluation may include precision on rare fake classes and error analysis of borderline cases.

Fake News Detection Lab — Bangalore

Classification, stance, fact-check and deployment setups for BE, BTech and MTech projects.

Fake vs Real
Classification

ISOT / Kaggle

Stance
Detection

Agree / Disagree

Claim
Verification

Evidence + NLI

Clickbait
Headlines

Pattern + ML

BERT
Fine-Tuning

Transformers

Feature
Engineering

Style + Source

Explainability
LIME / SHAP

Interpret

Viva & Report
Preparation

Documentation

Fake News Detection Projects.