Browsing by Subject "One-class classification"

Now showing 1 - 8 of 8

Open Access
Face manipulation detection
(2023-09) Nourmohammadi, Sepehr
Advancements in deep learning have facilitated the creation of highly realistic counterfeit human faces, ushering in the era of deepfakes. The potential to generate such convincingly authentic fake content prompts concerns due to the potential harm it could inflict on individuals and societies alike. Current studies predominantly focus on binary approaches that differentiate between real and fake images or videos. However, this approach can be time-consuming, requiring a multitude of diverse fake examples for training. Furthermore, unique deepfake content generated using different models may elude detection, making it challenging to apprehend all deepfakes. We propose two potential solutions. First, we suggest a one-class classification method, a purist approach that trains solely on real data and tests on both real and fake data. Second, using a cross-manipulation technique as a non-purist approach, which refers to the application of image manipulations to a use unseen/unknown manipulated samples during the training of the machine learning model. Efficacy in this process can be achieved by using a combination of different models, which enhances the detection of deep fakes. This is done by merging learning-based systems involving an ℓp-norm constraint with adjustable p-norm rules, thereby providing both sparse and non-sparse solutions to enhance discriminatory information between based learners in ensemble learning. Contrary to conventional subject-independent learning methods employed in deep fake detection, we propose a subject-dependent learning approach. Our preliminary findings suggest that this multifaceted approach can effectively detect deepfakes, demonstrating impressive results on the FaceForensics++ dataset as well as on generic one-class classification datasets including the UCI, and Keel datasets in both pure and non-pure approaches.
Open Access
Generalized texture models for detecting high-level structures in remotely sensed images
(2007) Doğrusöz, Emel
With the rapid increase in the amount and resolution of remotely sensed image data, automatic extraction and classification of information obtained from such images have been an important problem in the field of pattern recognition since remotely sensed imagery is a critical resource for diverse fields such as urban land use monitoring and management, GIS and mapping, environmental change and agricultural and ecological studies. This thesis proposes statistical and structural texture models for detecting high-level structures in remotely sensed images. The high-level structures correspond to complex geospatial objects with characteristic spatial layouts in a region. As opposed to the existing approaches that are based on classifying images using pixel level methods, we propose to use simple geospatial objects as textural primitives and exploit their spatial patterns. This representation can be viewed as a “generalized texture” measure where the image elements of interest are urban primitives instead of the traditional case of pixels. The spatial patterns we are interested in correspond to the regular and irregular arrangements of these primitives within neighborhoods. The methodology we propose in this thesis has two steps. First, the primitives of interest are detected using spectral, textural and morphological features with one-class classifiers. Then, the spatial patterns of these primitives are modeled. At this step, either a statistical or a structural approach can be followed. In the statistical approach, analysis of the spatial arrangement of the primitives is done by co-occurrence-based spatial domain features and Fourier spectrum-based frequency domain features. These features are used to quantify the likelihood of presence of the focused object in the image region being analyzed. In the structural approach, a graph-theoretic representation is proposed where the primitives form the nodes of a graph and the neighborhood information is obtained through Voronoi tessellation of the image scene. Next, the graph is clustered by thresholding its minimum spanning tree and the resulting clusters are classified as regular or irregular by examining the distributions of the angles between neighboring nodes. The algorithms proposed in this thesis are illustrated with the detection of two geospatial objects: settlement areas and harbors. The first step in the modeling of these objects is the detection of primitives such as buildings for settlement areas, and boats and water for harbors. In the second step, both statistical and structural approaches are illustrated for the modeling of the spatial patterns of these objects. Results of the experiments on high-resolution Ikonos satellite imagery and DOQQ aerial imagery show that the proposed techniques can be used for detecting the presence of geospatial objects in large remote sensing image datasets.
Embargo
lp-norm constrained one-class classifier combination
(Elsevier BV, 2024-02) Nourmohammadi, Sepehr; Rahimzadeh Arashloo, Shervin; Kittler, Josef
Classifier fusion is established as an effective methodology for boosting performance in different classification settings and one-class classification is no exception. In this study, we consider the one-class classifier fusion problem by modelling the sparsity/uniformity of the ensemble. To this end, we formulate a convex objective function to learn the weights in a linear ensemble model and impose a variable l(p >= 1)-norm constraint on the weight vector. The vector-norm constraint enables the model to adapt to the intrinsic uniformity/sparsity of the ensemble in the space of base learners and acts as a (soft) classifier selection mechanism by shaping the relative magnitudes of fusion weights. Drawing on the Frank-Wolfe algorithm, we then present an effective approach to solve the proposed convex constrained optimisation problem efficiently. We evaluate the proposed one-class classifier combination approach on multiple data sets from diverse application domains and illustrate its merits in comparison to the existing approaches.
Open Access
ℓp-norm support vector data description
(Elsevier BV, 2022-07-23) Arashloo, Shervin Rahimzadeh
The support vector data description (SVDD) approach serves as a de facto standard for one-class classification where the learning task entails inferring the smallest hyper-sphere to enclose target objects while linearly penalising the errors/slacks via an ℓ1-norm penalty term. In this study, we generalise this modelling formalism to a general ℓp-norm (p ≥ 1) penalty function on slacks. By virtue of an ℓp-norm function, in the primal space, the proposed approach enables formulating a non-linear cost for slacks. From a dual problem perspective, the proposed method introduces a dual norm into the objective function, thus, proving a controlling mechanism to tune into the intrinsic sparsity/uniformity of the problem for enhanced descriptive capability. A theoretical analysis based on Rademacher complexities characterises the generalisation performance of the proposed approach while the experimental results on several datasets confirm the merits of the proposed method compared to other alternatives.
Open Access
One-class classification using ℓp-norm multiple kernel fisher null approach
(Institute of Electrical and Electronics Engineers, 2023-03-14) Arashloo, Shervin Rahimzadeh
We address the one-class classification (OCC) problem and advocate a one-class MKL (multiple kernel learning) approach for this purpose. To this aim, based on the Fisher null-space OCC principle, we present a multiple kernel learning algorithm where an ℓp -norm regularisation ( p≥1 ) is considered for kernel weight learning. We cast the proposed one-class MKL problem as a min-max saddle point Lagrangian optimisation task and propose an efficient approach to optimise it. An extension of the proposed approach is also considered where several related one-class MKL tasks are learned concurrently by constraining them to share common weights for kernels. An extensive evaluation of the proposed MKL approach on a range of data sets from different application domains confirms its merits against the baseline and several other algorithms.
Open Access
Robust one-class classification using deep kernel spectral regression
(ELSEVIER, 2024-03-07) Mohammad, Salman; Arashloo, Shervin Rahimzadeh
The existing one-class classification (OCC) methods typically presume the existence of a pure target training set and generally face difficulties when the training set is contaminated with non-target objects. This work addresses this aspect of the OCC problem and formulates an effective method that leverages the advantages of kernel-based methods to achieve robustness against training label noise while enabling direct deep learning of features from the data to optimise a Fisher-based loss function in the Hilbert space. As such, the proposed OCC approach can be trained in an end-to-end fashion while, by virtue of a Tikhonov regularisation in the Hilbert space, it provides high robustness against the training set contamination. Extensive experiments conducted on multiple datasets in different application scenarios demonstrate that the proposed methodology is robust and performs better than the state-of-the-art algorithms for OCC when the training set is corrupted by contamination.
Open Access
Spoofing attack detection by anomaly detection
(Institute of Electrical and Electronics Engineers Inc., 2019) Fatemifar, S.; Arashloo, Shervin Rahimzadeh; Awais, M.; Kittler, J.
Spoofing attacks on biometric systems can seriously compromise their practical utility. In this paper we focus on face spoofing detection. The majority of papers on spoofing attack detection formulate the problem as a two or multiclass learning task, attempting to separate normal accesses from samples of different types of spoofing attacks. In this paper we adopt the anomaly detection approach proposed in [1], where the detector is trained on genuine accesses only using one-class classifiers and investigate the merit of subject specific solutions. We show experimentally that subject specific models are superior to the commonly used client independent method. We also demonstrate that the proposed approach is more robust than multiclass formulations to unseen attacks.
Open Access
Unseen face presentation attack detection using sparse multiple kernel fisher null-space
(IEEE, 2020) Arashloo, Shervin Rahimzadeh
We address the face presentation attack detection problem in the challenging conditions of an unseen attack scenario where the system is exposed to novel presentation attacks that were not available in the training stage. To this aim, we pose the unseen face presentation attack detection (PAD) problem as the one-class kernel Fisher null-space regression and present a new face PAD approach that only uses bona fide (genuine) samples for training. Drawing on the proposed kernel Fisher null-space face PAD method and motivated by the availability of multiple information sources, next, we propose a multiple kernel fusion anomaly detection approach to combine the complementary information provided by different views of the problem for improved detection performance. And the last but not the least, we introduce a sparse variant of our multiple kernel Fisher null-space face PAD approach to improve inference speed at the operational phase without compromising much on the detection performance. The results of an experimental evaluation on the OULU-NPU, Replay-Mobile, Replay-Attack and MSU-MFSD datasets illustrate that the proposed method outperforms other methods operating in an unseen attack detection scenario while achieving very competitive performance to multi-class methods (that benefit from presentation attack data for training) despite using only bona fide samples in the training stage.