Global phishing rose steadily from Q3 2024 to Q2 2025 (APWG), while smishing also increased in South Korea, leading KISA to launch the Cyber-Spider Project for early detection and attacker profiling.

Aligning text-to-image (T2I) diffusion models with preference optimization is valuable for human-annotated datasets, but the heavy cost of manual data collection limits scalability. Using reward models offers an alternative, however, current preference optimization methods fall short in exploiting the rich information, as they only consider pairwise preference distribution. Furthermore, they lack generalization to multi-preference scenarios and struggle to handle inconsistencies between rewards. To address this, we present Calibrated Preference Optimization (CaPO), a novel method to align T2I diffusion models by incorporating the general preference from multiple reward models without human annotated data. The core of our approach involves a reward calibration method to approximate the general preference by computing the expected win-rate against the samples generated by the pretrained models. Additionally, we propose a frontier-based pair selection method that effectively manages the multi-preference distribution by selecting pairs from Pareto frontiers. Finally, we use regression loss to fine-tune diffusion models to match the difference between calibrated rewards of a selected pair. Experimental results show that CaPO consistently outperforms prior methods, such as Direct Preference Optimization (DPO), in both single and multi-reward settings validated by evaluation on T2I benchmarks, including GenEval and T2I-Compbench.

Recent advances in diffusion models have introduced a new era of text-guided image manipulation, enabling users to create realistic edited images with simple textual prompts. However, there is significant concern about the potential misuse of these methods, especially in creating misleading or harmful content. Although recent defense strategies, which introduce imperceptible adversarial noise to induce model failure, have shown promise, they remain ineffective against more sophisticated manipulations, such as editing with a mask. In this work, we propose DiffusionGuard, a robust and effective defense method against unauthorized edits by diffusion-based image editing models, even in challenging setups. Through a detailed analysis of these models, we introduce a novel objective that generates adversarial noise targeting the early stage of the diffusion process. This approach significantly improves the efficiency and effectiveness of adversarial noises. We also introduce a mask-augmentation technique to enhance robustness against various masks during test time. Finally, we introduce a comprehensive benchmark designed to evaluate the effectiveness and robustness of methods in protecting against privacy threats in realistic scenarios. Through extensive experiments, we show that our method achieves stronger protection and improved mask robustness with lower computational costs compared to the strongest baseline. Additionally, our method exhibits superior transferability and better resilience to noise removal techniques compared to all baseline methods. Our source code is publicly available at https://choi403.github.io/diffusionguard.

Learning robust representations from data often requires scale, which has led to the success of recent zero-shot models such as CLIP. However, the obtained robustness can easily be deteriorated when these models are fine-tuned on other downstream tasks (e.g., of smaller scales). Previous works often interpret this phenomenon in the context of domain shift, developing fine-tuning methods that aim to preserve the original domain as much as possible. However, in a different context, fine-tuned models with limited data are also prone to learning features that are spurious to humans, such as background or texture. In this paper, we propose StarFT (Spurious Textual Alignment Regularization), a novel framework for fine-tuning zero-shot models to enhance robustness by preventing them from learning spuriosity. We introduce a regularization that aligns the output distribution for spuriosity-injected labels with the original zero-shot model, ensuring that the model is not induced to extract irrelevant features further from these descriptions. We leverage recent language models to get such spuriosity-injected labels by generating alternative textual descriptions that highlight potentially confounding features. Extensive experiments validate the robust generalization of StarFT and its emerging properties: zero-shot group robustness and improved zero-shot classification. Notably, StarFT boosts both worst-group and average accuracy by 14.30% and 3.02%, respectively, in the Waterbirds group shift scenario, where other robust fine-tuning baselines show even degraded performance.

An ultra-low power and tiny 8-bit SAR analogto-digital converter (ADC) for biomedical applications is proposed. The proposed SAR ADC introduces a hybrid capacitive-charge injection DAC (C-CI DAC) structure to achieve both high power efficiency and small area. Several methods to minimize the leakage current are used to reduce the power dissipation of the ADC operating at a low clock speed. The proposed C-CI SAR ADC is fabricated in a 28-nm CMOS process and operates from a 0.4-V supply voltage. At a sampling rate of 100 kS/s, the SNDR is 42.57 dB and the power consumption is 26.71 nW. The Walden FOM is 2.43 fJ/convstep and a core area of 389 um2is achieved.

This paper presents a high-speed CMOS image sensor (CIS) using a two-step analog-to-digital converter (ADC) that applies the self-successive doubling (SSD) algorithm. The proposed readout circuit uses a successive approximation register (SAR) ADC to implement a high-speed CIS and addresses the area requirements of the SAR ADC by employing an SSD circuit. The SSD circuit has a structure similar to conventional analog correlated double sampling (CDS) circuits, which simultaneously perform CDS and ADC operations. The proposed high-speed two-step ADC uses SSD logic for MSB and SAR ADC for LSB, thereby reducing the capacitance area by about 96.8% compared to a conventional 12-bit SAR-ADC. The proposed circuit is fabricated using a 180 nm process, with a total power consumption of 7.54 mW and a frame rate of 1190 fps.

Addressing biases in computer vision models is crucial for real-world AI deployments. However, mitigating visual biases is challenging due to their unexplainable nature, often identified indirectly through visualization or sample statistics, which necessitates additional human supervision for interpretation. To tackle this issue, we propose the Bias-to-Text (B2T) framework, which interprets visual biases as keywords. Specifically, we extract common keywords from the captions of mispredicted images to identify potential biases in the model. We then validate these keywords by measuring their similarity to the mispredicted images using a vision-language scoring model. The keyword explanation form of visual bias offers several advantages, such as a clear group naming for bias discovery and a natural extension for debiasing using these group names. Our experiments demonstrate that B2T can identify known biases, such as gender bias in CelebA, background bias in Waterbirds, and distribution shifts in ImageNet-R/C. Additionally, B2T uncovers novel biases in larger datasets, such as Dollar Street and ImageNet. For example, we discovered a contextual bias between “bee” and “flower” in ImageNet. We also highlight various applications of B2T keywords, including debiased training, CLIP prompting, and model comparison.11Code: https://github.com/alinlab/b2t

Recent progress in text-to-3D generation has been achieved through the utilization of score distillation methods: they make use of the pre-trained text-to-image (T2I) diffusion models by distilling via the diffusion model training objective. However, such an approach inevitably results in the use of random timesteps at each update, which increases the variance of the gradient and ultimately prolongs the optimization process. In this paper, we propose to enhance the text-to-3D optimization by leveraging the T2I diffusion prior in the generative sampling process with a predetermined timestep schedule. To this end, we interpret text-to-3D optimization as a multi-view image-to-image translation problem, and propose a solution by approximating the probability flow. By leveraging the proposed novel optimization algorithm, we design DreamFlow, a practical three-stage coarse-to-fine text-to-3D optimization framework that enables fast generation of high-quality and high-resolution (i.e., 1024×1024) 3D contents. For example, we demonstrate that DreamFlow is 5 times faster than the existing state-of-the-art text-to-3D method, while producing more photorealistic 3D contents.

Recently, various pre-training methods have been introduced in vision-based Reinforcement Learning (RL). However, their generalization ability remains unclear due to evaluations being limited to in-distribution environments and non-unified experimental setups. To address this, we introduce the Atari Pre-training Benchmark (Atari-PB), which pre-trains a ResNet-50 model on 10 million transitions from 50 Atari games and evaluates it across diverse environment distributions. Our experiments show that pre-training objectives focused on learning task-agnostic features (e.g., identifying objects and understanding temporal dynamics) enhance generalization across different environments. In contrast, objectives focused on learning task-specific knowledge (e.g., identifying agents and fitting reward functions) improve performance in environments similar to the pre-training dataset but not in varied ones. We publicize our codes, datasets, and model checkpoints at https://github.com/dojeon-ai/Atari-PB.

Text-to-image (T2I) diffusion models, when fine-tuned on a few personal images, can generate visuals with a high degree of consistency. However, such fine-tuned models are not robust; they often fail to compose with concepts of pretrained model or other fine-tuned models. To address this, we propose a novel fine-tuning objective, dubbed Direct Consistency Optimization, which controls the deviation between fine-tuning and pretrained models to retain the pretrained knowledge during fine-tuning. Through extensive experiments on subject and style customization, we demonstrate that our method positions itself on a superior Pareto frontier between subject (or style) consistency and image-text alignment over all previous baselines; it not only outperforms regular fine-tuning objective in image-text alignment, but also shows higher fidelity to the reference images than the method that fine-tunes with additional prior dataset. More importantly, the models fine-tuned with our method can be merged without interference, allowing us to generate custom subjects in a custom style by composing separately customized subject and style models. Notably, we show that our approach achieves better prompt fidelity and subject fidelity than those post-optimized for merging regular fine-tuned models.

Generative priors of large-scale text-to-image diffusion models enable a wide range of new generation and editing applications on diverse visual modalities. However, when adapting these priors to complex visual modalities, often represented as multiple images (e.g., video or 3D scene), achieving consistency across a set of images is challenging. In this paper, we address this challenge with a novel method, Collaborative Score Distillation (CSD). CSD is based on the Stein Variational Gradient Descent (SVGD). Specifically, we propose to consider multiple samples as “particles” in the SVGD update and combine their score functions to distill generative priors over a set of images synchronously. Thus, CSD facilitates the seamless integration of information across 2D images, leading to a consistent visual synthesis across multiple samples. We show the effectiveness of CSD in a variety of editing tasks, encompassing the visual editing of panorama images, videos, and 3D scenes. Our results underline the competency of CSD as a versatile method for enhancing inter-sample consistency, thereby broadening the applicability of text-to-image diffusion models.

Contrastive learning (CL) has been widely investigated with various learning mechanisms and achieves strong capability in learning representations of data in a self-supervised manner using unlabeled data. A common fashion of contrastive learning on this line is employing mega-sized encoders to achieve comparable performance as the supervised learning counterpart. Despite the success of the labelless training, current contrastive learning algorithms *failed* to achieve good performance with lightweight (compact) models, e.g., MobileNet, while the requirements of the heavy encoders impede the energy-efficient computation, especially for resource-constrained AI applications. Motivated by this, we propose a new self-supervised CL scheme, named SACL-XD, consisting of two technical components, **S**limmed **A**symmetrical **C**ontrastive **L**earning (SACL) and **Cross**-**D**istillation (XD), which collectively enable efficient CL with compact models. While relevant prior works employed a strong pre-trained model as the teacher of unsupervised knowledge distillation to a lightweight encoder, our proposed method trains CL models from scratch and outperforms them even without such an expensive requirement. Compared to the SoTA lightweight CL training (distillation) algorithms, SACL-XD achieves 1.79% ImageNet-1K accuracy improvement on MobileNet-V3 with 64$\times$ training FLOPs reduction.

Vision-language models, such as contrastive language-image pre-training (CLIP), have demonstrated impressive results in natural image domains. However, these models often struggle when applied to specialized domains like remote sensing, and adapting to such domains is challenging due to the limited number of image-text pairs available for training. To address this, we propose S-CLIP, a semi-supervised learning method for training CLIP that utilizes additional unpaired images. S-CLIP employs two pseudo-labeling strategies specifically designed for contrastive learning and the language modality. The caption-level pseudo-label is given by a combination of captions of paired images, obtained by solving an optimal transport problem between unpaired and paired images. The keyword-level pseudo-label is given by a keyword in the caption of the nearest paired image, trained through partial label learning that assumes a candidate set of labels for supervision instead of the exact one. By combining these objectives, S-CLIP significantly enhances the training of CLIP using only a few image-text pairs, as demonstrated in various specialist domains, including remote sensing, fashion, scientific figures, and comics. For instance, S-CLIP improves CLIP by 10% for zero-shot classification and 4% for image-text retrieval on the remote sensing benchmark, matching the performance of supervised CLIP while using three times fewer image-text pairs.

Knowledge distillation aims to transfer the information by minimizing the cross-entropy between the probabilistic outputs of the teacher and student network. In this work, we propose an alternative distillation objective by maximizing the scoring rule, which quantitatively measures the agreement of a distribution to the reference distribution. We demonstrate that the proper and homogeneous scoring rule exhibits more preferable properties for distillation than the original cross entropy based approach. To that end, we present an efficient implementation of the distillation objective based on a pseudo-spherical scoring rule, which is a family of proper and homogeneous scoring rules. We refer to it as pseudo-spherical knowledge distillation. Through experiments on various model compression tasks, we validate the effectiveness of our method by showing its superiority over the original knowledge distillation. Moreover, together with structural distillation methods such as contrastive representation distillation, we achieve state of the art results in CIFAR100 benchmarks.

Contrastive representation learning seeks to acquire useful representations by estimating the shared information between multiple views of data. Here, the choice of data augmentation is sensitive to the quality of learned representations: as harder the data augmentations are applied, the views share more task-relevant information, but also task-irrelevant one that can hinder the generalization capability of representation. Motivated by this, we present a new robust contrastive learning scheme, coined RényiCL, which can effectively manage harder augmentations by utilizing Rényi divergence. Our method is built upon the variational lower bound of a Rényi divergence, but a naive usage of a variational method exhibits unstable training due to the large variance. To tackle this challenge, we propose a novel contrastive objective that conducts variational estimation of a skew Renyi divergence and provides a theoretical guarantee on how variational estimation of skew divergence leads to stable training. We show that Rényi contrastive learning objectives perform innate hard negative sampling and easy positive sampling simultaneously so that it can selectively learn useful features and ignore nuisance features. Through experiments on ImageNet, we show that Rényi contrastive learning with stronger augmentations outperforms other self-supervised methods without extra regularization or computational overhead. Also, we validate our method on various domains such as graph and tabular datasets, showing empirical gain over original contrastive methods.

In this paper, we present a new on-device automatic speech recognition (ASR) system based on monotonic chunk-wise attention (MoChA) models trained with large (> 10K hours) corpus. We attained around 90% of a word recognition rate for general domain mainly by using joint training of connectionist temporal classifier (CTC) and cross entropy (CE) losses, minimum word error rate (MWER) training, layer-wise pretraining and data augmentation methods. In addition, we compressed our models by more than 3.4 times smaller using an iterative hyper low-rank approximation (LRA) method while minimizing the degradation in recognition accuracy. The memory footprint was further reduced with 8-bit quantization to bring down the final model size to lower than 39 MB. For on-demand adaptation, we fused the MoChA models with statistical n-gram models, and we could achieve a relatively 36% improvement on average in word error rate (WER) for target domains including the general domain.

In this paper, we present an end-to-end training framework for building state-of-the-art end-to-end speech recognition systems. Our training system utilizes a cluster of Central Processing Units (CPUs) and Graphics Processing Units (GPUs). The entire data reading, large scale data augmentation, neural network parameter updates are all performed “on-the-fly”. We use vocal tract length perturbation [1] and an acoustic simulator [2] for data augmentation. The processed features and labels are sent to the GPU cluster. The Horovod allreduce approach is employed to train neural network parameters. We evaluated the effectiveness of our system on the standard Librispeech corpus [3] and the 10,000-hr anonymized Bixby English dataset. Our end-to-end speech recognition system built using this training infrastructure showed a 2.44 % WER on test-clean of the LibriSpeech test set after applying shallow fusion with a Transformer language model (LM). For the proprietary English Bixby open domain test set, we obtained a WER of 7.92 % using a Bidirectional Full Attention (BFA) end-to-end model after applying shallow fusion with an RNN-LM. When the monotonic chunckwise attention (MoCha) based approach is employed for streaming speech recognition, we obtained a WER of 9.95 % on the same Bixby open domain test set.

The goal of differentially private data publishing is to release a modified dataset so that its privacy can be ensured while allowing for efficient learning. We propose a new data publishing algorithm in which a released dataset is formed by mixing ℓ randomly chosen data points and then perturbing them with an additive noise. Our privacy analysis shows that as ℓ increases, noise with smaller variance is sufficient to achieve a target privacy level. In order to quantify the usefulness of our algorithm, we adopt the accuracy of a predictive model trained with our synthetic dataset, which we call the utility of the dataset. By characterizing the utility of our dataset as a function of ℓ, we show that one can learn both linear and nonlinear predictive models so that they yield reasonably good prediction accuracies. Particularly, we show that there exists a sweet spot on ℓ that maximizes the prediction accuracy given a required privacy level, or vice versa. We also demonstrate that given a target privacy level, our datasets can achieve higher utility than other datasets generated with the existing data publishing algorithms.

This paper presents methods to accelerate recurrent neural network based language models (RNNLMs) for online speech recognition systems. Firstly, a lossy compression of the past hidden layer outputs (history vector) with caching is introduced in order to reduce the number of LM queries. Next, RNNLM computations are deployed in a CPU-GPU hybrid manner, which computes each layer of the model on a more advantageous platform. The added overhead by data exchanges between CPU and GPU is compensated through a frame-wise batching strategy. The performance of the proposed methods evaluated on LibriSpeech1test sets indicates that the reduction in history vector precision improves the average recognition speed by 1.23 times with minimum degradation in accuracy. On the other hand, the CPU-GPU hybrid parallelization enables RNNLM based real-time recognition with a four times improvement in speed.

In this study, the performance of a dual‐hop amplify‐and‐forward relaying system with beamforming is analysed, where only the source and destination are equipped with multiple antennas and both hops are subject to κ – μ fading channels. The κ – μ fading model is a general fading model that can accurately model practical small scale fading in line‐of‐sight environments and accommodates Rician, Nakagami‐ m , and Rayleigh as special cases. New exact analytical expressions on the outage probability (OP), average symbol error rate (SER), and average capacity are derived. Moreover, asymptotic results for the OP, SER, and average capacity are also derived in simpler forms in terms of basic elementary functions which make it easy to understand the system behaviour and the impact of the channel parameters. These analytical results are general and can emulate different symmetric and asymmetric fading scenarios as special cases such as Rician/Rician, Nakagami‐ m /Nakagami‐ m , Rayleigh/Rayleigh, and mixed κ – μ , Rician, Nakagami‐ m , and Rayleigh fading links.

Recurrent Neural Network Language Models (RNNLMs) have started to be used in various fields of speech recognition due to their outstanding performance. However, the high computational complexity of RNNLMs has been a hurdle in applying the RNNLM to a real-time Large Vocabulary Continuous Speech Recognition (LVCSR). In order to accelerate the speed of RNNLM-based network searches during decoding, we apply the General Purpose Graphic Processing Units (GPGPUs). This paper proposes a novel method of applying GPGPUs to RNNLM-based graph traversals. We have achieved our goal by reducing redundant computations on CPUs and amount of transfer between GPGPUs and CPUs. The proposed approach was evaluated on both WSJ corpus and in-house data. Experiments shows that the proposed approach achieves the real-time speed in various circumstances while maintaining the Word Error Rate (WER) to be relatively 10% lower than that of n-gram models.

Gaming on phones, tablets and laptops is very popular. Cloud gaming - where remote servers perform game execution and rendering on behalf of thin clients that simply send input and display output frames - promises any device the ability to play any game any time. Unfortunately, the reality is that wide-area network latencies are often prohibitive; cellular, Wi-Fi and even wired residential end host round trip times (RTTs) can exceed 100ms, a threshold above which many gamers tend to deem responsiveness unacceptable.

No abstract available.

Vehicular environments require continuous awareness of the road ahead. It is critical that mobile applications used in such environments (e.g., GPS route planners and location-based search) do not distract drivers from the primary task of operating the vehicle. Fortunately, a large body of research on vehicular interfaces provides best practices that mobile application developers can follow. However, when we studied the most popular vehicular applications in the Android marketplace, no application followed these guidelines. In fact, vehicular applications were not substantially better at meeting best practice guidelines than non-vehicular applications.

As more services have come to rely on sensor data such as audio and photos collected by mobile phone users, verifying the authenticity of this data has become critical for service correctness. At the same time, clients require the flexibility to tradeoff the fidelity of the data they contribute for resource efficiency or privacy. This paper describes YouProve, a partnership between a mobile device's trusted hardware and software that allows untrusted client applications to directly control the fidelity of data they upload and services to verify that the meaning of source data is preserved. The key to our approach is trusted analysis of derived data, which generates statements comparing the content of a derived data item to its source. Experiments with a prototype implementation for Android demonstrate that YouProve is feasible. Our photo analyzer is over 99% accurate at identifying regions changed only through meaning-preserving modifications such as cropping, compression, and scaling. Our audio analyzer is similarly accurate at detecting which sub-clips of a source audio clip are present in a derived version, even in the face of compression, normalization, splicing, and other modifications. Finally, performance and power costs are reasonable, with analyzers having little noticeable effect on interactive applications and CPU-intensive analysis completing asynchronously in under 70 seconds for 5-minute audio clips and under 30 seconds for 5-megapixel photos.

In IEEE 802.16j standard, relays are added to increase coverage and improve throughput. However, some overhead results from the inclusion of the relays. One typical example is an addition of R-MAP (relay MAP) to be included in the header of the frames transmitted from BS (base station) to RS (relay stations) in non-transparent mode [1]. The R-MAP conveys transmission schedules for the burst data from BS to an RS. Since R-MAP is mostly modulated by the lower MCS (Modulation and Coding Scheme) level (QPSK 1/2 for example), the bandwidth consumption is considerable. Compared with higher MCS level (64QAM 3/4 for example) used by data bursts, it uses 5 times the bandwidth. In this paper, we propose a scheme to compress the R-MAP and show its efficiency through simulation. We also show how to reduce the CID (connection identifiers) information in MAC headers to further reduce the overhead from relaying. Compression of R-MAP will greatly reduce the overhead from relaying. In other words, we reduce the overhead from relaying in control plane (R-MAP) as well as in data plane (MAC header). As confirmed by the performance evaluation, MAC efficiency of the proposed schemes is higher than that of the Standard schemes by approximately 11%.

A key feature of ubiquitous computing is service continuity which allows a user to transparently continue his task regardless of his movement. For service continuity, the underlying system needs to not only discover a service satisfying a user's request, but also provide an interface differences resolution scheme if the interface of the service found is not the same as that of the service requested. For resolving interface mismatches, one of solutions is to use an interface adapter. The most serious problem in the interface adapter-based approach is the overhead of adapter generation. There are many research efforts about adapter generation load reduction and this paper focuses on an adapter chaining scheme to reduce the number of necessary adapters among different service interfaces. We propose a construction-time adaptation loss evaluation scheme and an adapter chain construction algorithm, which finds an adapter chain with minimal adaptation loss.

This paper addresses the localization and navigation problem using invisible two dimensional barcodes on the floor. Compared with other methods using natural/artificial landmark, the proposed localization method has great advantages in cost and appearance, since the location of the robot is perfectly known using the barcode information after the mapping is finished. We also propose a navigation algorithm which uses the topological structure. For the topological information, we define nodes and edges which are suitable for indoor navigation, especially for large area having multiple rooms, many walls and many static obstacles. The proposed algorithm also has an advantage that errors occurred in each node are mutually independent and can be compensated exactly after some navigation using barcode. Simulation and experimental results were performed to verify the algorithm in the barcode environment, and the result showed an excellent performance. After mapping, it is also possible to solve the kidnapped case and generate paths using topological information

This paper proposes and evaluates a scalable dynamic load distribution scheme for multi-server distributed virtual environment systems, where users are highly skewed rather than uniformly distributed over a virtual environment. In the proposed scheme, an overloaded server initiating load distribution selects a set of servers to be involved in load distribution by dynamically adapting to the workload status of other servers, unlike the existing approaches. Upon completion of server selection, the intiating server repartitions the regions dedicated to the involved servers using a graph partitioning algorithm so that all the involved servers have the roughly equal workload. The involved servers then migrate their workloads with each other in a peer-to-peer manner according to the result of repartitioning. We present and analyze the simulation results that compare the performance of the proposed scheme with that of the existing schemes.