FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments
Open accessThe DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index; 6 of the 14 sources its claims cite were checked for support.
Abstract
The widespread deployment of CCTV systems has significantly enhanced surveillance and public safety across various environments. However, the emergence of deepfake technology poses serious challenges by enabling malicious manipulation of video footage, compromising the reliability of CCTV systems for evidence collection and privacy protection. Existing deepfake detection solutions often suffer from high computational overhead and are unsuitable for real-time deployment on resource-constrained CCTV cameras. This paper proposes FL-TENB4, a Federated-Learning-enhanced Tiny EfficientNetB4-Lite framework for deepfake detection in CCTV environments. The proposed architecture integrates Tiny Machine Learning (TinyML) techniques with EfficientNetB4-Lite, a lightweight convolutional neural network optimized for edge devices, and employs a Federated Learning (FL) approach for collaborative model updates. The TinyML-based local model ensures real-time deepfake detection with minimal latency, while FL enables privacy-preserving training by aggregating model updates without transferring sensitive video data to centralized servers. The effectiveness of the proposed system is validated using the FaceForensics++ dataset under resource-constrained conditions. Experimental results demonstrate that FL-TENB4 achieves high detection accuracy, reduced model size, and low inference latency, making it highly suitable for real-world CCTV environments.
Overview
CoChat Summary
## TL;DR
FL-TENB4 is a federated-learning-enhanced TinyML deepfake detector built on EfficientNetB4-Lite for resource-constrained CCTV environments. On FaceForensics++ it reports 94.2% accuracy, 93.5% F1-score, 0.96 ROC-AUC, 12 ms per frame latency, 4.2 MB model size, and 150 mW inference energy.
## Key findings
- On FaceForensics++ (reported as C0, C23, and C40), FL-TENB4 achieved 96.5% accuracy on C0, 94.2% on C23, and 89.8% on C40; the paper says C23 is most representative of real CCTV conditions.
- The model reported 94.2% detection accuracy, 93.5% F1-score, 0.96 ROC-AUC, 12 ms per frame inference latency, 4.2 MB model size, and 150 mW energy consumption during inference.
- Compared with listed baselines, FL-TENB4 outperformed Multi-View Inconsistency (91.8%) and CNN-SVM (90.2%), and was slightly above or comparable to XceptionNet (93.5%) in accuracy.
- The paper states that FL-TENB4 reduces bandwidth requirements through quantization/sparsification/selective aggregation and supports scalable collaborative training across edge devices.
## Methods
- The paper proposes a three-layer architecture: device layer, edge layer, and cloud layer, combining TinyML with Federated Learning (FedAvg) for privacy-preserving deepfake detection in CCTV video streams.
- The local detector is EfficientNetB4-Lite optimized with quantization and pruning; frames are resized to 260 × 260, normalized, converted to tensors, and classified as real/fake in real time.
- Experiments/simulations used FaceForensics++ and in-house CCTV data; the paper reports a 70:15:15 train/validation/test split and evaluation across C0, C23, and C40 compression levels, with comparisons to several existing methods.
## Limitations
- The paper states that reliance on the FaceForensics++ dataset and a custom CCTV dataset constrains generalizability to real-world scenarios and a broader range of video qualities, compression levels, and manipulation types.
- The paper notes that real-world CCTV footage often contains frames without human faces, while the model focuses on facial manipulations; it proposes a face-detection module as a remedy.
- The text also makes evident that the FL framework does not explicitly address adversarial attacks such as model poisoning during aggregation; this is stated in the discussion, not as a formal limitation section.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures





Verification
Claims the paper makes
4 Valid · 6 Not valid · 8 Not checked · 86 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Additionally, these systems often lack real-time capabilities, making them impractical for dynamic surveillance operations [8].”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 1. Introduction
What the sources say
- Partly supportedcites [8] Yu, 2021, A survey on deepfake video detection, IET Biom. sourceExperimental — accuracy not yet measured
Section 5.1.3 (Time consumption; page unavailable) says current detection algorithms are too slow for wide use in practical scenarios and discusses their prospective use on streaming media platforms. It does not specifically address dynamic surveillance operations, so that application is an unsupported extension. (LLM, conf=0.86)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Most state-of-the-art models require extensive computational resources, rendering them unsuitable for resource-constrained environments such as edge devices integrated into CCTV cameras.”
- Also stated in: § Abstract · plain
“Existing deepfake detection solutions often suffer from high computational overhead and are unsuitable for real-time deployment on resource-constrained CCTV cameras.”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 1. Introduction
What the sources say
- Partly supportedcites [8] Yu, 2021, A survey on deepfake video detection, IET Biom. sourceExperimental — accuracy not yet measured
The Discussion (section 5; page unavailable) says current research tends toward complex networks that increase calculation costs. Time consumption (section 5.1.3; page unavailable) says existing algorithms are too slow for wide use in practical scenarios. The paper does not establish that most state-of-the-art models are unsuitable specifically for resource-constrained CCTV edge devices. (LLM, conf=0.87)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Abstract · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Despite these fluctuations, the average latency remains low and consistent”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 4. Simulation and Analysis of the Proposed Architecture > 4.2. Analysis
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Experimental results demonstrate that FL-TENB4 achieves high detection accuracy, reduced model size, and low inference latency, making it highly suitable for real-world CCTV environments.”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > Abstract
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For the C40 dataset, accuracy decreased to 89.8% due to the significant visual detail loss caused by heavy compression.”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 4. Simulation and Analysis of the Proposed Architecture > 4.2. Analysis
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
34 more claims
Create a free account to see all 104 claims and every source verdict.
- It may state two separate claims as one. 44
- Probably not a substantial claim. 42
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 3. Proposed Architecture > 3.3. FL-Based Global and Local Model Update to Enhance Deepfake Detection Performance
“In this step, the server computes a weighted average of the local model parameters based on the size of the local datasets on each device.”
§ FL-TENB4: A Federated-Learning-Enhanced Tiny EfficientNetB4-Lite Approach for Deepfake Detection in CCTV Environments > 3. Proposed Architecture > 3.3. FL-Based Global and Local Model Update to Enhance Deepfake Detection Performance
“The updated global model wt+1 is computed as follows:(7)wt+1← 1N∗∑i=1Nwit where N is the total number of participating edge devices and wit are the parameters of the local model after training on device i.”
Citations
Cites
FaceForensics++: Learning to Detect Manipulated Facial Images
Rossler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., & Niessner, M. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00009 · Cited by 3,314 · cited 2×
“[13] Rössler et al.”
A robust privacy-preserving federated learning model against model poisoning attacks
Yazdinejad, A., Dehghantanha, A., Karimipour, H., Srivastava, G., & Parizi, R. M. (2024) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2024.3420126 · Cited by 380 · cited 3×
“Abbas Yazdinejad et al. [22] introduced an FL model designed to counteract model poisoning attacks using Gaussian Mixture Models (GMM) and Mahalanobis Distance (MD) for anomaly detection in gradient updates, coupled with Additive Homomorphic Encryption (AHE) for data confidentiality.”
InfluentialDeepFake Detection for Human Face Images and Videos: A Survey
Malik, A., Kuribayashi, M., Abdullahi, S. M., & Khan, A. N. (2022) · IEEE Access · doi:10.1109/access.2022.3151186 · Cited by 297 · cited 1×
“This capability is critical for maintaining the integrity of surveillance footage and enabling prompt responses to potential security threats [28].”
A Survey on Deepfake Video Detection
Yu, P., Xia, Z., Fei, J., & Lu, Y. (2021) · IET Biometrics · doi:10.1049/bme2.12031 · Cited by 253
Block Hunter: Federated Learning for Cyber Threat Hunting in Blockchain-Based IIoT Networks
Yazdinejad, A., Dehghantanha, A., Parizi, R. M., Hammoudeh, M., Karimipour, H., & Srivastava, G. (2022) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2022.3168011 · Cited by 201
Deepfake generation and detection: Case study and challenges
Patel, Y., Tanwar, S., Gupta, R., Bhattacharya, P., Davidson, I. E., Nyameko, R., Aluvala, S., & Vimal, V. (2023) · IEEE Access · doi:10.1109/access.2023.3342107 · Cited by 186 · cited 1×
“As a result, the system becomes highly feasible for deployment in real-world CCTV environments, even on devices with minimal computational resources [30].”
Deepfake Detection through Deep Learning
Pan, D., Sun, L., Wang, R., Zhang, X., & Sinnott, R. O. (2020) · 2020 IEEE/ACM International Conference on Big Data Computing, Applications and Technologies (BDCAT) · doi:10.1109/bdcat50828.2020.00001 · Cited by 172
FAMM: Facial Muscle Motions for Detecting Compressed Deepfake Videos Over Social Networks
Liao, X., Wang, Y., Wang, T., Hu, J., & Wu, X. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3278310 · Cited by 169 · cited 2×
“[19] Xin Liao et al.”
A Data Verification System for CCTV Surveillance Cameras Using Blockchain Technology in Smart Cities
Khan, P., Byun, Y.-C., & Park, N. (2020) · Electronics · doi:10.3390/electronics9030484 · Cited by 144
DeepfakeStack: A Deep Ensemble-based Learning Technique for Deepfake Detection
Rana, M. S., & Sung, A. H. (2020) · 2020 7th IEEE International Conference on Cyber Security and Cloud Computing (CSCloud)/2020 6th IEEE International Conference on Edge Computing and Scalable Cloud (EdgeCom) · doi:10.1109/cscloud-edgecom49738.2020.00021 · Cited by 133 · cited 1×
“Rana et al. [14] proposed DeepfakeStack, an ensemble-based learning technique that combines multiple deep neural networks to improve deepfake detection performance.”
Snis: a signal noise separation-based network for post-processed image forgery detection
Chen, J., Liao, X., Wang, W., Qian, Z., Qin, Z., & Wang, Y. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3204753 · Cited by 129
Deepfake Attacks: Generation, Detection, Datasets, Challenges, and Research Directions
Naitali, A., Ridouani, M., Salahdine, F., & Kaabouch, N. (2023) · Computers · doi:10.3390/computers12100216 · Cited by 117 · cited 1×
“These refinements ensure superior detection performance in resource-constrained environments, making EfficientNetB4-Lite an ideal solution for deployment on edge devices without compromising performance [27].”
Federated Quantum-Based Privacy-Preserving Threat Detection Model for Consumer Internet of Things
Namakshenas, D., Yazdinejad, A., Dehghantanha, A., & Srivastava, G. (2024) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2024.3377550 · Cited by 110
A Hybrid CNN-LSTM Approach for Precision Deepfake Image Detection Based on Transfer Learning
Al-Dulaimi, O. A. H. H., & Kurnaz, S. (2024) · Electronics · doi:10.3390/electronics13091662 · Cited by 91
Namakshenas, D., Yazdinejad, A., Dehghantanha, A., Parizi, R. M., & Srivastava, G. (2024) · IEEE Transactions on Industrial Cyber-Physical Systems · doi:10.1109/ticps.2024.3435178 · Cited by 82
A Contemporary Survey on Deepfake Detection: Datasets, Algorithms, and Challenges
Gong, L. Y., & Li, X. J. (2024) · Electronics · doi:10.3390/electronics13030585 · Cited by 75 · cited 1×
“This approach eliminates the need for extensive infrastructure and centralized computation, making it highly adaptable to diverse and distributed CCTV systems [31].”
A Review of Deepfake Techniques: Architecture, Detection and Datasets
Edwards, P., Nebel, J.-C., Greenhill, D., & Liang, X. (2024) · IEEE Access · doi:10.1109/access.2024.3477257 · Cited by 50
Interactive Two-Stream Network Across Modalities for Deepfake Detection
Wu, J., Zhang, B., Li, Z., Pang, G., Teng, Z., & Fan, J. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3269841 · Cited by 46
Gotcha: Real-Time Video Deepfake Detection via Challenge-Response
Mittal, G., Hegde, C., & Memon, N. D. (2024) · IEEE European Symposium on Security and Privacy · doi:10.1109/eurosp60621.2024.00009 · Cited by 42
Deepfake Image Detection using CNNs and Transfer Learning
Kumar, N., Pranav, P., Nirney, V., & Geetha, V. (2021) · International Conference on Computing, Communication and Green Engineering (CCGE) · doi:10.1109/ccge50943.2021.9776410 · Cited by 39
Deepfake Detection and Localization Using Multi-View Inconsistency Measurement
Zhang, B., Yin, Q., Lu, W., & Luo, X. (2024) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2024.3472064 · Cited by 35
Capsule-Forensics Networks for Deepfake Detection
Nguyen, H. H., Yamagishi, J., & Echizen, I. (2022) · Advances in computer vision and pattern recognition · doi:10.1007/978-3-030-87664-7_13 · Cited by 30
Welsh, B. C., Piza, E. L., Thomas, A. L., & Farrington, D. P. (2019) · Journal of Contemporary Criminal Justice · doi:10.1177/1043986219890192 · Cited by 29
Amount-Based Covert Communication Over Blockchain
Tian, Y., Liao, X., Dong, L., Xu, Y., & Jiang, H. (2024) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2024.3358013 · Cited by 21
CSTAN: A Deepfake Detection Network with CST Attention for Superior Generalization
Yang, R., You, K., Pang, C., Luo, X., & Lan, R. (2024) · Sensors · doi:10.3390/s24227101 · Cited by 12
Deepfake Detection Via Separable Self-Consistency Learning
Lu, L., Wang, Y., Zhuo, W., Zhang, L., Gao, G., & Guo, Y. (2024) · Proceedings - International Conference on Image Processing · doi:10.1109/icip51287.2024.10647481 · Cited by 5
Detection of Deepfake Images and Videos Using SVM, CNN, and Hybrid Approaches
Stankov, I. S., & Dulgerov, E. E. (2024) · International Scientific Conference Electronics (ET) · doi:10.1109/et63133.2024.10721497 · Cited by 5
Analysis of the Reliability of Deepfake Facial Emotion Expression Synthesis
López, J. M., Gil Iranzo, R. M., & García González, R. (2024) · Repositori ObertUDL (University of Lleida) · doi:10.22967/hcis.2024.14.035 · Cited by 1
Research works printed but not linked: 3
- Jeremiah, 2024, PrivacyGuard: Collaborative Edge-Cloud Computing Architecture for Attribute-Preserving Face Anonymization in CCTV Networks, Hum. Centric Comput. Inf. Sci., 14, 1
- Castro, 2023, Comprehensive Survey on AI-Based Technologies for Enhancing IoT Privacy and Security: Trends, Challenges, and Solutions, Hum. Centric Comput. Inf. Sci., 13, 39
- Kim, 2023, MBTI Personality Type Prediction Model Using WZT Analysis Based on the CNN Ensemble and GAN, Hum. Centric Comput. Inf. Sci., 13, 14
Cited by
Federated Learning for Surveillance Systems: A Literature Review and AHP Expert-Based Evaluation
Shin, Y.-S., Kim, H., Jeong, J., & Shin, D. (2025) · Electronics · doi:10.3390/electronics14173500 · Cited by 4
Reddy, P. K., & Rao, K. S. (2026) · Research Square · doi:10.21203/rs.3.rs-10197752/v1
Cognitive-Inspired Multimodal Deepfake Detection Using Privacy-Preserving Federated Intelligence
Blfgeh, A., & Ardah, H. (2026) · Cognitive Computation · doi:10.1007/s12559-026-10640-4 · Cited by 1
Privacy Preserving Distributed Image Processing Using Federated Learning and CNNs
Ranasinghe, N. K., Liyanage, P., & Kruglova, L. V. (2025) · IEEE Symposium on Computer Applications & Industrial Electronics (ISCAIE) · doi:10.1109/iscaie64985.2025.11080797 · Cited by 3
FedMaxAdapt: Gradient-Guided Adaptive Sparsity for Scalable Multi-Modal Federated Deepfake Detection
Manzoor, H. U., Javed, A., Mir, J., Yousaf, M. H., Manzoor, S., & Zoha, A. (2026) · International Conference on Robotics and Automation in Industry (ICRAI) · doi:10.1109/icrai70912.2026.11551901 · Cited by 1
Metadata
- Authors
- Ha, J., El Azzaoui, A., & Park, J. H.
- Affiliations
- Seoul National University of Science and Technology (KR)
- Department of Computer Science and Engineering, Seoul National University of Science and Technology, Seoul 01811, Republic of Korea
- Corresponding author
- Park, Jong Hyuk
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2025-01-28
- Volume
- 25
- Issue
- 3
- Pages
- 788
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC11821191
- OpenAlex
- W4406908477
- Cited by
- 13
- References
- 27
- Citations by year
- 2026: 9 · 2025: 4
- FWCI
- 4.52
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Korean government (2780000007)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Keywords
- deepfake detection, federated learning, TinyML, CCTV surveillance, edge computing, lightweight neural networks, privacy-preserving training, real-time video analysis, FaceForensics++
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.