Deepfake Audio Detection via MFCC Features Using Machine Learning

IEEE Access, Journal Year: 2022, Volume and Issue: 10, P. 134018 - 134028

Published: Jan. 1, 2022

Deepfake content is created or altered synthetically using artificial intelligence (AI) approaches to appear real. It can include synthesizing audio, video, images, and text. Deepfakes may now produce natural-looking content, making them harder identify. Much progress has been achieved in identifying video deepfakes recent years; nevertheless, most investigations detecting audio have employed the ASVSpoof AVSpoof dataset various machine learning, deep learning algorithms. This research uses learning-based identify deepfake audio. Mel-frequency cepstral coefficients (MFCCs) technique used acquire useful information from We choose Fake-or-Real dataset, which benchmark dataset. The was with a text-to-speech model divided into four sub-datasets: for-rece, for-2-sec, for-norm for-original. These datasets are classified sub-datasets mentioned above according length bit rate. experimental results show that support vector (SVM) outperformed other (ML) models terms of accuracy on for-rece for-2-sec datasets, while gradient boosting performed very well VGG-16 produced highly encouraging when applied for-original outperforms state-of-the-art approaches.

Language: Английский

---

A Comprehensive Survey on Rare Event Prediction

ACM Computing Surveys, Journal Year: 2024, Volume and Issue: 57(3), P. 1 - 39

Published: Oct. 14, 2024

Rare event prediction involves identifying and forecasting events with a low probability using machine learning (ML) data analysis. Due to the imbalanced distributions, where frequency of common vastly outweighs that rare events, it requires specialized methods within each step ML pipeline, is, from processing algorithms evaluation protocols. Predicting occurrences is important for real-world applications, such as Industry 4.0, an active research area in statistical ML. This article comprehensively reviews current approaches along four dimensions: data, processing, algorithmic approaches, approaches. Specifically, we consider 73 datasets different modalities (i.e., numerical, image, text, audio), major categories five groupings, two broader aims identify gaps literature highlight challenges predicting events. It also suggests potential directions, which can help guide practitioners researchers.

Language: Английский

---

Data Augmentation-based Novel Deep Learning Method for Deepfaked Images Detection

ACM Transactions on Multimedia Computing Communications and Applications, Journal Year: 2023, Volume and Issue: 20(11), P. 1 - 15

Published: April 13, 2023

Recent advances in artificial intelligence have led to deepfake images, enabling users replace a real face with genuine one. images recently been used malign public figures, politicians, and even average citizens. but realistic stir political dissatisfaction, blackmail, propagate false news, carry out bogus terrorist attacks. Thus, identifying from fakes has got more challenging. To avoid these issues, this study employs transfer learning data augmentation technique classify images. For experimentation, 190,335 RGB-resolution image methods are prepare the dataset. The experiments use deep models: convolutional neural network (CNN), Inception V3, visual geometry group (VGG19), VGG16 approach. Essential evaluation metrics (accuracy, precision, recall, F1-score, confusion matrix, AUC-ROC curve score) test efficacy of proposed Results revealed that approach achieves an accuracy, F1-score score 90% 91% our fine-tuned model outperforming other DL models recognizing deepfakes.

Language: Английский

---

Deepfake audio detection via MFCC features and mel-spectrogram using deep learning

AIP conference proceedings, Journal Year: 2025, Volume and Issue: 3264, P. 030027 - 030027

Published: Jan. 1, 2025

Language: Английский

---

Real-time active-learning method for audio-based anomalous event identification and rare events classification for audio events detection

Frontiers in Computer Science, Journal Year: 2025, Volume and Issue: 7

Published: April 28, 2025

Introduction Audio event detection, the application of scientific methods to analyze audio recordings, can be helpful in examining and analyzing recordings preserve, analyze, interpret sound evidence. Furthermore, it safety compliance, security, surveillance, maintenance, predictive analysis. detection aims recover meaningful information from such as determining authenticity recording, identifying speakers, reconstructing conversations. However, filtering out noise for better accuracy is a major challenge. A greater sense public security achieved by developing automated systems that are both cost-effective real-time. Methods In response these challenges, this study presented method anomalous events based on noisy evidence real-time scenario help investigator during investigation. This created large dataset containing original audio. The includes diverse environmental background settings (e.g., office, restaurant, park) some abnormal explosion, car crash, human attack). used an ensemble learning model conduct experiments active environment. Nine employed create feature vector. Results show proposed using obtained score 99.26%, while deep 95.35%. was tested scenario. Discussion experiment results approach efficiently detect

Language: Английский

---

Geolocation from Images: A New Frontier in Cybersecurity Applications

Lecture notes in electrical engineering, Journal Year: 2025, Volume and Issue: unknown, P. 390 - 401

Published: Jan. 1, 2025

Language: Английский

---

An Analysis of Abnormal Event Detection and Person Identification from Surveillance Cameras using Motion Vectors with Deep Learning

Published: March 2, 2023

In many places, like museums, banks, and airports, network cameras have taken the place of outdated analogue as a result advancements in chip technology reduced costs storage bandwidth equipment, among other factors. an endeavor to increase public safety health protection, lower crime, implement video surveillance, sector has entered "blowout" phase. A crucial part developing intelligent CCTV system is detection abnormal events, human behaviour, object recognition. These technologies enable anomalous environmental phenomena, state alert. systems use machine learning vision capabilities recognize detect specific anomalies that appear stream footage. For these systems, supervised popular training method, frame-by-frame processing frequently used. However, been replaced unsupervised well semi-supervised for system's since come variety forms because it not practical which was before educate all types anomalies. By using this technology, amount labour must be done by humans manually spot live feed generate alerts can or eliminated. Additionally, method improves efficiency only preserving occurrences their original quality leaving regular circumstances low quality. you may Grassmann algorithm identify faces data alert surrounding security systems. The suggested approach offers better reliability outlier face detection, according experimental findings.

Language: Английский

---

A review of deep learning techniques in audio event recognition (AER) applications

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(3), P. 8129 - 8143

Published: June 14, 2023

Language: Английский

---

Multimedia datasets for anomaly detection: a review

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(19), P. 56785 - 56835

Published: Dec. 13, 2023

Language: Английский

---

Evaluating the Role of Data Enrichment Approaches towards Rare Event Analysis in Manufacturing

Sensors, Journal Year: 2024, Volume and Issue: 24(15), P. 5009 - 5009

Published: Aug. 2, 2024

Rare events are occurrences that take place with a significantly lower frequency than more common, regular events. These can be categorized into distinct categories, from frequently rare to extremely rare, based on factors like the distribution of data and significant differences in rarity levels. In manufacturing domains, predicting such is particularly important, as they lead unplanned downtime, shortening equipment lifespans, high energy consumption. Usually, inversely correlated maturity industry. Typically, affects multivariate generated within process highly imbalanced, which leads bias predictive models. This paper evaluates role enrichment techniques combined supervised machine learning for event detection prediction. We use time series augmentation sampling address scarcity, maintaining its patterns, imputation handle null values. Evaluating 15 models, we find improves F1 measure by up 48% Our empirical ablation experiments provide novel insights, also investigate model interpretability.

Language: Английский

---