Search Results (68)

Vehicle theft and vandalism remain significant urban security challenges commonly addressed through reactive, post-incident forensic measures. This paper proposes a proactive, real-time computer vision system designed to detect potentially suspicious behavior around parked vehicles, with a specific focus on unauthorized proximity and loitering. The proposed architecture integrates state-of-the-art object detection using YOLOv11 (You Only Look Once version 11), multi-object tracking via a lightweight custom association tracker inspired by the ByteTrack/StrongSORT/OC-SORT paradigm, and monocular depth estimation based on the Intel DPT-Large framework.A key contribution is the identification and mitigation of the Perspective Challenge: the two-dimensional (2D) scale ambiguity that causes distant background pedestrians to appear falsely proximate to foreground vehicles in monocular camera feeds. To address this, three spatial analysis strategies are implemented and evaluated: (A) fixed Euclidean thresholding, (B) adaptive perspective thresholding, and (C) three-dimensional (3D) depth injection. Experimental results on real-world urban surveillance footage (27,000 annotated frames across two datasets) demonstrate that Strategy C achieves the highest precision (0.95) with an F1-score of 0.92, while Strategy B provides the best balance between accuracy (precision 0.88, recall 0.91, F1 0.89) and computational efficiency (32.7 frames per second, FPS). Compared to naive 2D thresholding (Strategy A), Strategy B reduces false alarms by approximately 80%, while Strategy C further improves precision to 0.95 through depth-plane verification. The system maintains real-time performance exceeding 30 FPS under Strategy B, making it a strong candidate for practical urban vehicle monitoring, subject to further large-scale validation across diverse environments. Full article

Deepfakes threaten sensor-based authentication systems, including biometric sensors, surveillance cameras, and IoT edge devices. Unimodal detectors remain vulnerable to modality-specific attacks. We propose a multimodal deepfake detection framework optimized for resource-constrained edge devices, featuring a novel cross-modal attention fusion mechanism with adaptive gating. The architecture combines enhanced Res2Net for audio, temporal 3D CNN with SE attention for video, and bidirectional cross-modal attention with quality-based gates. On our benchmark (5472 audio + 1842 video samples), the fusion model achieves 96.7% accuracy, 96.6% F1-score, 0.988 AUC-ROC, and 3.3% EER. Adversarial testing shows 92.3% accuracy under the Fast Gradient Sign Method (FGSM) attack. The model has a 30.3 MB footprint and runs at 20 FPS on edge hardware. Modality contribution analysis reveals adaptive weighting (72% audio for TTS forgery, 78% video for lip-synced attacks). Cross-dataset evaluation on FakeAVCeleb achieves 92.3% overall accuracy, confirming generalization. Full article

Pedestrian–vehicle collisions produce a rich kinematic record that is entirely lost by the time a forensic investigation begins. Recovering this record constitutes a state-estimation problem. This paper presents a Phase 1 design for a multimodal sensor fusion and signal-processing framework utilising 128-channel LiDAR, 1080p NIR stereo cameras, and a 2 kHz IMU, all fused via Kalman filtering and Savitzky–Golay polynomial differentiation. The framework is evaluated through Monte Carlo uncertainty propagation and sensitivity analysis applied to a constructed simulation scenario; no real clinical or forensic data are used in this Phase 1 report. Under simulated conditions with throw-distance measurement uncertainty of ±0.5 m, velocity reconstruction shows an estimated propagated uncertainty of ±2.03 km/h under expanded simulation conditions with vehicle-coefficient variance activated. Sensitivity analysis indicates that a 10% noise spike in acceleration would theoretically amplify injury metrics by 26.9%, providing quantitative justification for noise-optimal pre-filtering. The bimodal kinematic–acoustic architecture is proposed as a physically interpretable foundation for collision reconstruction; its experimental performance awaits Phase 2–4 validation. A five-phase validation roadmap is presented, progressing from FEA simulation to independent multi-site replication before any forensic deployment is proposed. Full article

Analog digital video recorders (DVRs) are still extensively used in small-to-medium business and home security systems, but there are special problems when it comes to forensic recovery of video evidence in these systems that are not covered by tools or methodology. Compared to the IP-based network video recorders, analog DVRs packetize video frames of several coaxial-connected cameras into a single interleaved binary stream on disk, necessitating channel demultiplexing before single camera footage can be reassembled. In this paper, we discuss a multi-channel analog Dahua DVR system utilizing the DHAV frame format, with a focus on the forensic recovery approach. Three significant contributions are presented in the methodology: (1) a channel demultiplexing algorithm that separates interleaved frames with up to 32 cameras on the basis of embedded channel identifiers and temporal coherence analysis; (2) a frame sequence stitching mechanism to reassemble continuous video segments on the basis of non-contiguous disk fragments using adaptive frame number tolerance (±3 frames) and temporal validation (≤1 second difference); and (3) a native C implementation with Win32 GUI providing significant performance improvements over interpreted alternatives. The system was tested on 14 analog Dahua DVR hard drives of various models, with a 92.3% recovery rate (97.1% on hard drives with no hardware damage), 91.3% temporal accuracy, 97.5% channel separation accuracy and a 1.8% false positive rate. The methodology fills an important gap in the literature of surveillance forensics, where current studies have only concentrated on IP-based digital systems, and analog DVRs form an estimated 35–40% of operational surveillance systems across emerging markets. The channel demultiplexing capability, which is not found in any current commercial or academic tool, enables automated per-camera organization of interleaved streams, converting what was previously a manual multi-day process into an automated one. Full article

Biometric systems are widely used in security and forensic applications. Conventionally, contact-based footprint scanners require physical contact, which presents significant limitations. These devices raise hygiene concerns and are impractical in field identification conditions, such as forensic investigations or disaster victim identification, where quick and non-invasive methods are essential. To address these challenges, a contactless footprint acquisition and identification system was developed using image processing techniques and a Convolutional Neural Network (CNN) based on the Visual Geometry Group–16 layer architecture. The system employs a Raspberry Pi 4, a Logitech C922 camera, and a ring light to capture footprints without direct surface contact. Captured images are processed with Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve contrast and mean thresholding to generate binary images for clearer feature extraction. System performance was evaluated using a multiclass confusion matrix. The CNN correctly classified 158 of 160 test images, achieving an accuracy of 98.75%. This result demonstrates higher accuracy than earlier studies that used older CNN models, such as Alex Krizhevsky’s Network and LeCun’s Network-5, which performed with fewer subjects and lower accuracy rates. The developed system shows potential for biometric security, forensic investigations, and disaster response, where contactless and reliable identification is required. Future research can expand the dataset with more diverse footprints, test performance under varied conditions, and extend the approach to other contactless biometrics such as palmprints or ears. Full article

The rapid deployment of surveillance cameras in urban, industrial, and domestic environments has intensified the need for intelligent systems capable of analyzing video streams beyond the limitations of single-camera setups. Unlike traditional single-camera approaches, multi-camera systems expand spatial coverage, reduce blind spots, and enable consistent tracking of people and objects across non-overlapping views, thereby improving robustness against occlusions and viewpoint changes. This article presents a comprehensive review of multi-camera vision systems published between 2020 and 2025, covering application domains including public security and biometrics, intelligent transportation, smart cities and IoT, healthcare monitoring, precision agriculture, industry and robotics, pan–tilt–zoom (PTZ) camera networks, and emerging areas such as retail and forensic analysis. The review synthesizes predominant technical approaches, including deep-learning-based detection, multi-target multi-camera tracking (MTMCT), re-identification (Re-ID), spatiotemporal fusion, and edge computing architectures. Persistent challenges are identified, particularly in inter-camera data association, scalability, computational efficiency, privacy preservation, and dataset availability. Emerging trends such as distributed edge AI, cooperative camera networks, and active perception are discussed to outline future research directions toward scalable, privacy-aware, and intelligent multi-camera infrastructures. Full article

Source camera identification on smartphones constitutes a fundamental task in multimedia forensics, providing essential support for applications such as image copyright protection, illegal content tracking, and digital evidence verification. Numerous techniques have been developed for this task over the past decades. Among existing approaches, Photo-Response Non-Uniformity (PRNU) has been widely recognized as a reliable device-specific fingerprint and has demonstrated remarkable performance in real-world applications. Nevertheless, the rapid advancement of computational photography technologies has introduced significant challenges: modern devices often exhibit anomalous behaviors under PRNU-based analysis. For instance, images captured by different devices may exhibit unexpected correlations, while images captured by the same device can vary substantially in their PRNU patterns. Current approaches are incapable of automatically exploring the underlying causes of these anomalous behaviors. To address this limitation, we propose a simple yet effective forensic analysis framework leveraging Exchangeable Image File Format (EXIF) metadata. Specifically, we represent EXIF metadata as type-aware word embeddings to preserve contextual information across tags. This design enables visual interpretation of the model’s decision-making process and provides complementary insights for identifying the anomalous behaviors observed in modern devices. Extensive experiments conducted on three public benchmark datasets demonstrate that the proposed method not only achieves state-of-the-art performance for source camera identification but also provides valuable insights into anomalous device behaviors. Full article

In this paper, we propose a source camera identification method based on disentangled feature modeling, aiming to achieve robust extraction of camera fingerprint features under complex imaging and post-processing conditions. To address the severe coupling between image content and camera fingerprint features in existing methods, which makes content interference difficult to suppress, we develop a dual-branch deep learning framework guided by imaging physics. By introducing physical consistency constraints, the proposed framework explicitly separates image content representations from device-related fingerprint features in the feature space, thereby enhancing the stability and robustness of source camera identification. The proposed method adopts two parallel branches: a content modeling branch and a fingerprint feature extraction branch. The content branch is built upon an improved U-Net architecture to reconstruct scene and color information, and further incorporates texture refinement and multi-scale feature fusion to reduce residual content interference in fingerprint modeling. The fingerprint branch employs ResNet-50 as the backbone network to learn discriminative global features associated with the camera imaging pipeline. Based on these branches, fingerprint information dominated by sensor noise is explicitly extracted by computing the residual between the input image and the reconstructed content, and is further encoded through noise analysis and feature fusion for joint camera model classification. Experimental results on multiple public-source camera forensics datasets demonstrate that the proposed method achieves stable and competitive identification performance in same-brand camera discrimination, complex imaging conditions, and post-processing scenarios, validating the effectiveness of the proposed disentangled modeling and physical consistency constraint strategy for source camera identification. Full article

The merging of the Internet of Things (IoT) and Artificial Intelligence (AI) advances has intensified challenges related to data authenticity and security. These advancements necessitate a multi-layered security approach to ensure the security, reliability, and integrity of critical infrastructure and intelligent surveillance systems. This paper proposes a two-layered security approach that combines a discrete cosine transform least significant bit 2 (DCT-LSB-2) with artificial neural networks (ANNs) for data forensic validation and mitigating deepfakes. The proposed model encodes validation codes within the LSBs of cover images captured by an IoT camera on the sender side, leveraging the DCT approach to enhance the resilience against steganalysis. On the receiver side, a reverse DCT-LSB-2 process decodes the embedded validation code, which is subjected to authenticity verification by a pre-trained ANN model. The ANN validates the integrity of the decoded code and ensures that only device-originated, untampered images are accepted. The proposed framework achieved an average SSIM of 0.9927 across the entire investigated embedding capacity, ranging from 0 to 1.988 bpp. DCT-LSB-2 showed a stable Peak Signal-to-Noise Ratio (average 42.44 dB) under various evaluated payloads ranging from 0 to 100 kB. The proposed model achieved a resilient and robust multi-layered data forensic validation system. Full article

Automotive dashboard cameras are widely used to record driving events and often serve as critical evidence in accident investigations and insurance claims. However, the availability of free and low-cost editing tools has increased the risk of video tampering, underscoring the need for reliable methods to verify video authenticity. Temporal tampering typically involves manipulating frame order through insertion, deletion, or duplication. This paper proposes a computationally efficient framework that transforms high-dimensional video into compact one-dimensional temporal signals and learns tampering patterns using a shallow one-dimensional convolutional neural network (1D-CNN). Five complementary features are extracted between consecutive frames: frame-difference magnitude, structural similarity drift (SSIM drift), optical-flow mean, forward–backward optical-flow consistency error, and compression-aware temporal prediction error. Per-video robust normalization is applied to emphasize intra-video anomalies. Experiments on a custom dataset derived from D²-City demonstrate strong detection performance in single-attack settings: 95.0% accuracy for frame deletion, 100.0% for frame insertion, and 95.0% for frame duplication. In a four-class setting (non-tampered, insertion, deletion, duplication), the model achieves 96.3% accuracy, with AUCs of 0.994, 1.000, 0.997, and 0.988, respectively. Efficiency analysis confirms near real-time CPU inference (≈12.7–12.9 FPS) with minimal memory overhead. Cross-dataset tests on BDDA and VIRAT reveal domain-shift sensitivity, particularly for deletion and duplication, highlighting the need for domain adaptation and augmentation. Overall, the proposed multi-feature 1D-CNN provides a practical, interpretable, and resource-aware solution for temporal tampering detection in dashcam videos, supporting trustworthy video forensics in IoT-enabled transportation systems. Full article

Photo-response non-uniformity (PRNU) has long been regarded as a reliable method for source camera identification and device linking in forensic applications. Recent advances in deep learning (DL) have introduced diverse architectures, including convolutional neural networks, residual learning, encoder–decoder representations, dual-branch structures, and contrastive learning, to capture specific sensor artifacts. This review summarizes the performance of these DL techniques across both tasks and compares their effectiveness at the model and device levels over time. While DL approaches achieve strong model-level accuracy, robust device-level identification remains challenging, particularly in modern imaging pipelines that involve camera-integrated or AI-driven enhancements during capture. These findings underscore the need for improved techniques and updated datasets to address evolving photograph capture practices. Full article

This review explores the current landscape of neural network-based methods for digital image noise processing. Digital cameras have become ubiquitous in fields like forensics and medical diagnostics, and image noise remains a critical factor for ensuring image quality. Traditional noise suppression techniques are often limited by extensive parameter selection and inefficient handling of complex data. In contrast, neural networks, particularly convolutional neural networks, autoencoders, and generative adversarial networks, have shown significant promise for noise estimation, suppression, and analysis. These networks can handle complex noise patterns, leverage context-specific data, and adapt to evolving conditions with minimal manual intervention. This paper describes the basics of camera and image noise components and existing techniques for their evaluation. Main neural network-based methods for noise estimation are briefly presented. This paper discusses neural network application for noise suppression, classification, image source identification, and the extraction of unique camera fingerprints through photo response non-uniformity. Additionally, it highlights the challenges of generating reliable training datasets and separating image noise from photosensor noise, which remains a fundamental issue. Full article

Several approaches are currently being used by law enforcement to locate the remains of victims. Yet, traditional methods are invasive and time-consuming. Unmanned Aerial Vehicle (UAV)-based remote sensing has emerged as a potential tool to support the location of human remains and clandestine graves. While offering a non-invasive and low-cost alternative, UAV-based remote sensing needs to be tested and validated for forensic case work. To assess current knowledge, a systematic review of 19 peer-reviewed articles from four databases was conducted, focusing specifically on UAV-based remote sensing for human remains and clandestine grave location. The findings indicate that different sensors (colour, thermal, and multispectral cameras), were tested across a range of burial conditions and models (human and mammalian). While UAVs with imaging sensors can locate graves and decomposition-related anomalies, experimental designs from the reviewed studies lacked robustness in terms of replication and consistency across models. Trends also highlight the potential of automated detection of anomalies over manual inspection, potentially leading to improved predictive modelling. Overall, UAV-based remote sensing shows considerable promise for enhancing the efficiency of human remains and clandestine grave location, but methodological limitations must be addressed to ensure findings are relevant to real-world forensic cases. Full article

Unmanned Aerial Vehicles (UAVs) have become increasingly valuable for accident scene reconstruction and forensic surveying due to their flexibility and ability to capture high-resolution imagery. This study investigates the impact of flight altitude, camera angle, and image overlap on the spatial accuracy of 3D models generated from UAV imagery. A total of 27 flight configurations were conducted using a DJI Phantom 4 Pro V2, combining three altitudes (30 m, 45 m, 60 m), three camera angles (90°, 75°, 60°), and three overlap levels (60%, 70%, 80%). The resulting 3D models were assessed by comparing measured linear distances between ground control points with known reference distances. The Root Mean Square Error (RMSE) was used to quantify model accuracy. The results indicated that lower flight altitudes, nadir or moderately oblique camera angles, and higher image overlaps consistently yielded the most accurate reconstructions. A Wilcoxon rank-sum test confirmed that the differences in accuracy across parameter settings were statistically significant. These findings highlight the critical role of flight configuration in achieving centimeter-level accuracy, as evidenced by RMSE values ranging from 1.7 to 7.6 cm, and provide practical recommendations for optimizing UAV missions in forensic and engineering applications. Full article

Digital dentistry is undergoing rapid evolution, with three-dimensional imaging technologies increasingly integrated into routine clinical workflows. Originally developed for accurate dental arch reconstruction, modern intraoral scanners have demonstrated expanding versatility in capturing intraoral mucosal as well as perioral cutaneous structures. Concurrently, photogrammetry has emerged as a powerful method for full-face digital reconstruction, particularly valuable in orthodontic and prosthodontic treatment planning. These advances offer promising applications in forensic sciences, where high-resolution, three-dimensional documentation of anatomical details such as palatal rugae, lip prints, and bite marks can provide objective and enduring records for legal and investigative purposes. This study explores the forensic potential of two digital acquisition techniques by presenting two cadaveric cases of animal bite injuries. In the first case, an intraoral scanner (Dexis 3600) was used in an unconventional extraoral application to directly scan skin lesions. In the second case, photogrammetry was employed using a digital single-lens reflex (DSLR) camera and Agisoft Metashape, with standardized lighting and metric scale references to generate accurate 3D models. Both methods produced analyzable digital reconstructions suitable for forensic archiving. The intraoral scanner yielded dimensionally accurate models, with strong agreement with manual measurements, though limited by difficulties in capturing complex surface morphology. Photogrammetry, meanwhile, allowed for broader contextual reconstruction with high texture fidelity, albeit requiring more extensive processing and scale calibration. A notable advantage common to both techniques is the avoidance of physical contact and impression materials, which can compress and distort soft tissues, an especially relevant concern when documenting transient evidence like bite marks. These results suggest that both technologies, despite their different origins and operational workflows, can contribute meaningfully to forensic documentation of bite-related injuries. While constrained by the exploratory nature and small sample size of this study, the findings support the viability of digitized, non-destructive evidence preservation. Future perspectives may include the integration of artificial intelligence to assist with morphological matching and the establishment of digital forensic databases for pattern comparison and expert review. Full article