FIDA (Feature Instability-Driven Attack), a novel backdoor attack framework that uses subtle semantic triggers for injection, but its key innovation is a novel objective called Feature Instability Loss, which effectively evades the evaluated perturbation-based defenses.
Abstract
Self-supervised learning (SSL) models are vulnerable to backdoor attacks. However, the systemic risks they pose in face representation have received little attention. The entanglement of identity features in self-supervised face learning presents unique challenges for attack stealthiness. To address this gap, we propose FIDA (Feature Instability-Driven Attack), a novel backdoor attack framework. FIDA uses subtle semantic triggers for injection, but its key innovation is a novel objective called Feature Instability Loss. It trains the encoder to increase the sensitivity of triggered features along perturbation directions sampled during attack optimization . By preventing the backdoor from exhibiting the rigid feature patterns typical of previous attacks, FIDA effectively evades the evaluated perturbation-based defenses. Experiments show that FIDA achieves a high attack success rate and generally preserves benign utility across the evaluated settings , posing a significant threat to real-world multimedia applications relying on facial analysis.
Due to the presence of semantic ambiguity among similar expression categories and the inherent imbalance in spatio-temporal feature intensities, dynamic facial expression recognition (DFER) in the wild poses significant challenges for building trustworthy and robust systems. These factors often lead to inconsistent feature representations and unreliable decision boundaries, which hinder the model’s ability to perform stable and accurate recognition under uncertainty, and further pose a serious safety hazard, e.g., misdiagnosis of depression. To tackle these challenges, we propose a novel adaptive framework, Semantic-Aware Facial Expression Recognition framework (SAFE), which is developed from an Information Bottleneck (IB)-inspired perspective to improve the robustness and prediction reliability of DFER in complex, unconstrained scenarios. Specifically, we first design a Temporal-aware Augmentation Module (TAM) to introduce structurally perturbed yet temporally coherent training samples, effectively mitigating spatio-temporal feature imbalance. Then, to ensure stable long-range modeling under temporal variation, we introduce the Spatio-temporal Modeling Module (STM) with a sparsity-aware state-space fusion gate. Furthermore, an Ambiguity-aware Calibration Loss (ACL) is formulated to dynamically refine decision boundaries by focusing on confusing and underrepresented categories, improving the model’s resilience to distributional skew and semantic uncertainty. Extensive experiments on two large-scale in-the-wild DFER benchmarks, DFEW and FERV39k, demonstrate that SAFE consistently outperforms state-of-the-art methods across multiple metrics, particularly under ambiguous and imbalanced conditions. These results validate the effectiveness of our approach in promoting more robust and stable expression recognition, which is important for trustworthy DFER in real-world environments. Codes are released at https://github.com/QIcita/SAFE_DFER
Feng-Qi Cui, Anyang Tong, Jinyang Huang et al.· IEEE Transactions on Informa...· 0 citations
This work proposes SRAP, which combines per-channel truncated SVD refinement with an identity-importance mask at every optimization step, and demonstrates that SRAP substantially improves protected-image fidelity across all reported metrics while maintaining competitive identity-disruption performance.
The proposed DiffAttack framework significantly outperforms existing adversarial techniques, achieving a high average attack success rate of 84.86% across multiple face recognition models (e.g., FaceNet).
Omid Ahmadieh, Nima Karimian· arXiv.org· 0 citations
Facial biometric recognition systems currently face compound threats intertwining generative AI and high-fidelity physical spoofing. Existing defenses suffer from systemic bottlenecks, including poor generalization, non-auditable reasoning, and reliance on massive, low-quality datasets. To address these challenges, we propose Multimodal Large Language Models (MFAD) for face anti-spoofing detection, an explainable reasoning system for Unified Face Anti-Spoofing Detection (UFAD), accompanied by a semantic-level annotation benchmark. Unlike methods relying on external tools or coarse alignment, MFAD activates the intrinsic reasoning capabilities of Multimodal Large Language Models (MLLMs) via a fine-grained pixel-semantic anchoring mechanism. This eliminates localization hallucinations and ensures auditable reasoning paths. We introduce a cross-attack semantic-level unified annotation paradigm: by annotating only 1,000 precise masks per attack category, we generate reasoning evidence chains strictly corresponding to spoofed regions. Supervised fine-tuning on the Qwen-VL foundation model demonstrates that, using limited high-quality samples, the system achieves a 40-50% relative reduction in in-domain ACER and restricts cross-domain performance degradation to within 11.62%/5.23%, significantly outperforming existing frameworks. Furthermore, under white-box adversarial attacks, detection accuracy drops by only 3.2%, validating the robustness of semantic anchoring compared to models trained on massive short-text data. Domain practitioners rated the evidence reliability of reasoning paths at 4.57/5, with inference latency satisfying real-time deployment requirements. These results confirm that a few-shot, high-quality semantic annotation paradigm is effective for building trustworthy, explainable, and cost-efficient UFAD systems.
Xiaoyong Yu, Rongzhen Li, Shuming Shi et al.· 0 citations
The reliability of finger vein biometric systems is increasingly threatened by sophisticated presentation attacks. Current presentation attack detection (PAD) methods, often relying on supervised learning, are vulnerable to novel, unseen attacks because they depend on comprehensive labeled spoof datasets that are impractical to collect. To address this zero-day threat, this research proposes a one-class anomaly detection framework trained solely on authentic finger vein samples. While utilizing a standard U-Net-inspired denoising autoencoder as the architectural backbone, this work introduces two key novel contributions to tailor the model for biometric security: (1) a Multi-Objective Anomaly Detection Loss Framework that uniquely integrates multi-scale reconstruction error, gradient preservation constraints, and deep Support Vector Data Description loss to strictly regularize the latent space and (2) a Spectral Error Optimization technique that applies adaptive frequency weighting to amplify subtle texture artifacts inherent in spoof mediums. This combination is significant because the multi-objective loss forces the model to learn fine-grained physiological vein patterns, while the spectral optimization captures high-frequency anomalies often missed by spatial reconstruction alone. Experimental results on the FVPAD-USM dataset demonstrate that this approach achieves an Attack Presentation Classification Error Rate below 2% and an Average Classification Error Rate below 10%. By outperforming supervised baselines like support vector machines and convolutional neural networks in generalizing to unseen digital and glossy photo attacks, this work establishes the effectiveness of unsupervised, frequency-enhanced anomaly detection for robust biometric security.
Received: 30 August 2025 | Revised: 18 March 2026 | Accepted: 18 June 2026
Conflicts of Interest
The authors declare that they have no conflicts of interest to this work.
Data Availability Statement
The data that support the findings of this study are openly available in the FVPAD-USM database at http://drfendi.com/fvpad_usm_database/, reference number [33].
Author Contribution Statement
Mohd Shahrimie Mohd Asaari: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft. Bakhtiar Affendi Rosdi: Validation, Investigation, Writing review & editing, Supervision. Andrew Tiong Hoe Pin: Methodology, Software, Validation, Investigation, Data curation, Visualization. Zahid Ur Rahman: Investigation, Resources, Data curation. Muhammad Firdaus Akbar: Validation, Funding acquisition.
M. Asaari, B. Rosdi, Andrew Tiong Hoe Pin et al.· Artificial Intelligence and...· 0 citations
Facial expression recognition (FER) is a challenging task that involves inferring human emotions from facial images entangled with various attributes. Although FER models based on Transformer have recently reported impressive performance, two major types of bias continue to degrade accuracy—namely, age-related facial attributes such as wrinkles and skin texture, which introduce confusion in recognizing elderly emotions, and imbalanced training data, which limits the model's ability to generalize across age groups. This paper proposes a novel bias-mitigation method that decouples and regularizes age- and emotion-related components within the self-attention mechanism of Transformer. The proposed method separately regularizes the value vectors that encode texture information correlated with age, and the query-key matrices that focus on facial landmarks crucial for emotion recognition. It encourages intra-class compactness of facial landmarks for emotion representation while minimizing age interference in features unrelated to emotion. To suppress age information and preserve emotional features, an age discriminator is employed to guide the value vectors in eliminating age cues, while an emotion classifier restores discriminative information for classification. In addition, the proposed method leverages triplet learning on the query-key space to enhance age-invariant emotion separation. Experiments conducted on four widely used FER benchmarks demonstrate that our method notably reduces age-related bias while maintaining or exceeding state-of-the-art performance.
Jaeil Park, Sung-Bae Cho· International Journal of Int...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
