Deep learning systems processing temporal and sequential data are increasingly deployed in safety-critical applications including healthcare monitoring, autonomous navigation, and algorithmic trading. However, these systems exhibit severe vulnerabilities to adversarial attacks—carefully crafted perturbations that cause systematic misclassification while remaining imperceptible. This paper presents a comprehensive systematic survey of adversarial attacks on time series classification, human activity recognition (HAR), and reinforcement learning (RL) systems, reviewing 127 papers published between 2019 and 2025 following PRISMA guidelines with documented inter-rater reliability (kappa = 0.83).We establish a unified four-dimensional taxonomy distinguishing attack characteristics across target modalities (wearable IMU sensors, WiFi/radar sensing, skeleton-based recognition, medical/financial time series, and RL agents), perturbation strategies, temporal scope, and physical realizability levels. Our quantitative synthesis reveals severe baseline vulnerabilities—FGSM attacks degrade HAR accuracy from 95.1% to 3.4% under white-box conditions—while demonstrating that cross-sensor transferability varies dramatically from 0% to 80% depending on body placement and modality. Critically, we identify a substantial gap between digital attack success rates (85–98%) and physically validated attacks, with hardware-in-the-loop validation demonstrating 70–97% success only for WiFi and radar modalities, while wearable IMU physical attacks remain entirely unvalidated.We provide systematic analysis of defense mechanisms including adversarial training, detection-based approaches, certified defenses, and ensemble methods, proposing the Temporal AutoAttack (T-AutoAttack) framework for standardized adaptive attack evaluation. Our analysis reveals that current defenses exhibit 6–23% performance degradation under adaptive attacks, with certified methods showing the smallest gap but incurring 15–30% clean accuracy costs. We further identify emerging vulnerabilities in transformer-based HAR architectures and LLM-based time series forecasters that require urgent attention.The survey culminates in a prioritized research roadmap identifying eight critical gaps with specific datasets, evaluation pipelines, and implementation timelines. We provide actionable deployment recommendations for practitioners across wearable HAR, WiFi/radar sensing, RL systems, and emerging LLM-based temporal applications. This work offers the first unified treatment bridging time series and reinforcement learning adversarial research, establishing foundations for developing robust temporal AI systems suitable for real-world deployment in safety-critical domains.

Nighttime vehicle detection poses significant challenges due to reduced visibility, uneven illumination, and increased noise in low-light imagery. While deep learning approaches have achieved remarkable success in daytime scenarios, their application to nighttime conditions remains constrained by the scarcity of specialized datasets and the computational demands of existing architectures. This paper presents three primary contributions to address these challenges. First, we introduce the Low-light Vehicle Annotation Dataset (L-VAD), comprising 13,648 annotated frames captured exclusively during nighttime conditions across three vehicle categories: motorcycle, car, and truck/bus. Second, we propose TinyNight-YOLO, an ultra-lightweight detection architecture achieving competitive performance with only ∼1.0 million parameters—representing a 2.6× reduction compared to YOLO11- N and 26.4× reduction compared to YOLO11-L. Third, we provide a comprehensive benchmark evaluating ten model variants across YOLO11 and YOLOv12 families. Experimental results demonstrate that TinyNight-YOLO achieves F1-Score of 0.9207 and mAP@50 of 0.9474, representing only 1.44% accuracy reduction compared to models 2.6× larger, while outperforming YOLOv12-L (26.4M parameters) despite having 26.4× fewer parameters. Among full-scale models, YOLO11-L achieves the highest F1-Score (0.9486), while YOLO11-M attains superior mAP@50-95 (0.7271). The L-VAD dataset is publicly available at Mendeley Data (doi: 10.17632/h6p2w53my5.1), providing the research community with a dedicated resource for advancing nighttime vehicle detection. The proposed TinyNight-YOLO architecture enables practical deployment on resource-constrained edge devices while maintaining detection accuracy above 94% mAP@50.

Deep learning-based Human Activity Recognition (HAR) systems using multimodal wearable sensors are increasingly deployed in safety-critical applications including healthcare monitoring, elderly care, and security authentication. However, the vulnerability of these systems to adversarial attacks remains insufficiently understood, particularly for attacks that must evade detection while manipulating multiple sensor modalities simultaneously. This paper presents STAR-RL (Stealth-aware Targeted Adversarial attack via Reinforcement Learning), a novel framework that generates effective and stealthy adversarial examples against multimodal sensor-based HAR systems. STAR-RL introduces three key innovations: (1) a multi-strategy attack engine that adaptively selects among diverse perturbation algorithms based on real-time attack progress, (2) a sensor-aware stealth mechanism that concentrates perturbations on naturally noisy sensors to minimize detection likelihood, and (3) a reinforcement learning-based meta-controller that learns optimal attack policies through interaction with the target classifier. Comprehensive experiments on the MHEALTH dataset demonstrate that STAR-RL achieves 95.20% attack success rate, substantially outperforming baseline methods including FGSM (6.00%), PGD (88.60%), and C&W (69.00%). The stealth analysis confirms that 51.35% of perturbation energy is successfully directed to weak sensors (gyroscopes and magnetometers), validating the effectiveness of the sensor-aware allocation strategy. Our findings reveal critical security vulnerabilities in production HAR systems and provide insights for developing robust defense mechanisms against adaptive adversarial threats.

Deep learning-based human activity recognition (HAR) systems, despite achieving high accuracy, remain vulnerable to adversarial attacks that pose severe threats to safety-critical deployments. This paper presents a comprehensive framework for sensor-specific targeted adversarial attacks on multimodal HAR systems. We propose a hybrid optimization strategy combining momentum-based Projected Gradient Descent with adaptive Carlini-Wagner optimization, incorporating dynamic early stopping and intelligent fallback mechanisms. Our approach constrains perturbations to individual sensor modalities, enabling systematic vulnerability assessment across heterogeneous configurations. Through extensive evaluation on the MHealth dataset with 96 sensor-target combinations and 38,000+ adversarial examples, our hybrid strategy achieves 96.46% targeted attack success rate—representing 45% improvement over baseline C&W and 8% over enhanced PGD—while maintaining 49× computational efficiency. Analysis reveals accelerometers exhibit highest vulnerability (99.83%), followed by gyroscopes (96.67-99.00%) and magnetometers (91.00-95.50%). High-motion activities prove universally vulnerable (100%), while sedentary activities show sensor-dependent robustness (66-100%). Statistical validation confirms strong correlation between model confidence and vulnerability (r = 0.71, p < 0.01). Limited cross-sensor transferability (28-42%) suggests promising defense directions through sensor redundancy and ensemble methods. Our findings underscore urgent needs for adversarially robust HAR design in safety-critical applications.

ChatGPT, sebuah model bahasa bertenaga AI yang dikembangkan oleh OpenAI, memiliki potensi untuk merevolusi proses pembelajaran di sektor pendidikan. ChatGPT dapat membantu para guru mengembangkan strategi pembelajaran yang lebih efektif dan efisien, terutama dalam konteks Sekolah Menengah Atas (SMA). Penerapan ChatGPT dalam pendidikan dapat menghasilkan konsep pendidikan baru, meningkatkan hasil belajar mengajar, dan menciptakan pengalaman belajar yang interaktif dan menarik. Hal ini juga menyediakan sumber daya pendidikan yang relevan, meningkatkan konten pendidikan. Namun, penerapan ChatGPT dapat menimbulkan bias dalam memahami konteks dan budaya lokal, sehingga sangat penting bagi guru untuk secara aktif menggunakan dan memvalidasi materi yang disediakan oleh ChatGPT. Hasilnya, ChatGPT dapat secara signifikan meningkatkan efektivitas dan kualitas perencanaan pendidikan, meskipun masih diperlukan intervensi dan bimbingan dari guru untuk memastikan keakuratan dan keefektifan materi pembelajaran

Deep neural networks have become the dominant approach to wearable human activity recognition (HAR), yet their opacity creates practical barriers: clinicians cannot verify that predictions align with physiological knowledge, and engineers lack principled guidance on which sensors to retain when cost and power constraints demand simplification. Existing explainable AI (XAI) methods for HAR typically rely on a single attribution technique, risking method-specific biases, and report importance at granularities individual channels or entire body locations that do not map cleanly onto hardware design choices. This paper introduces a multi-method XAI framework that addresses these gaps by systematically combining counterfactual sensor-group ablation, Integrated Gradients (IG), and Shapley Value Sampling around a common deep learning backbone. We evaluate the framework on a Time-Distributed LSTM trained on the MHEALTH dataset, which records twelve activities via eight sensor groups (accelerometers, gyroscopes, magnetometers, and ECG) at the chest, ankle, and wrist; the model achieves 98.2% accuracy and macro-F1 of 0.98. Four coordinated experiments probe model behaviour at global, sensor-group, and per-class levels. Counterfactual ablation reveals that removing the ankle magnetometer collapses accuracy for dynamic locomotion by 47.1 percentage points, while wrist accelerometer removal erodes confidence for static postures by more than 50%. Class-specific IG heatmaps and temporal attribution curves expose activity-dependent sensor signatures consistent with biomechanical expectations: ankle-centric patterns for gait, wrist- and chest-centric patterns for upper-body movements, and diffuse low-energy profiles for resting states. Global IG and Shapley rankings converge, with accelerometers accounting for 89% of total attribution mass. The agreement across causal, gradient-based, and game-theoretic perspectives strengthens confidence that identified importance patterns reflect genuine model behaviour. Together, these sensor-group-level explanations provide actionable guidance for sensor selection, power-aware deployment, and clinically meaningful interpretation without sacrificing recognition performance.