Toxic speech detection has become a crucial challenge in maintaining safe online communication environments. However, existing approaches to toxic speech detection often neglect the contribution of paralinguistic cues, such as emotion, intonation, and speech rate, which are key to detecting speech toxicity. Moreover, current toxic speech datasets are predominantly text-based, limiting the development of models that can capture paralinguistic cues. To address these challenges, we present ToxiAlert-Bench, a large-scale audio dataset comprising over 30,000 audio clips annotated with seven major toxic categories and twenty fine-grained toxic labels. Uniquely, our dataset annotates toxicity sources—distinguishing between textual content and paralinguistic origins—for comprehensive toxic speech analysis. Furthermore, we propose a dual-head neural network with a multi-stage training strategy tailored for toxic speech detection. This architecture features two task-specific classification headers: one for identifying the source of sensitivity (textual or paralinguistic), and the other for categorizing the specific toxic type. The training process involves independent head training followed by joint fine-tuning to reduce task interference. To mitigate data class imbalance, we incorporate class-balanced sampling and weighted loss functions. Our experimental results show that leveraging paralinguistic features significantly improves detection performance. Our method consistently outperforms existing baselines across multiple evaluation metrics, with a 21.1% relative improvement in Macro-F1 score and a 13.0% relative gain in accuracy over the strongest baseline, highlighting its enhanced effectiveness and practical applicability.

Symbolic execution is a powerful technique that can accurately synthesize program inputs for program testing through constraint solving. Applying symbolic execution effectively means that we must solve two searching problems efficiently. One is to search through the many program paths and the other is, given a particular path condition, to search through the numerous variable assignments to identify one satisfying solution. With few exceptions, existing symbolic execution engines treat constraint solvers as black boxes. As a result, the two searches are completely separated, which results in much redundancy (i.e., the same variable assignments may be tried for solving many program paths). Existing attempts on addressing this issue include those approaches based on constrained Horn clauses (in which the whole program is encoded as one constraint) and one preliminary attempt on caching and reusing partial solving results from the constraint solver. In this work, we propose SEC , which systematically computes the reward of concretizing a program path (for symbolic execution) and a variable (for constraint solving) and uses the reward as guide for integrating the two searches. We implemented SEC based on KLEE and evaluated it on a diverse set of programs. The results show that SEC is effective, i.e., achieving 15% more code coverage than the state-of-the-art baseline symbolic execution engines. Furthermore, we show that SEC can be readily combined with a state-of-the-art concolic testing engine to improve its performance

Voice communication is convenient while also poses risks of privacy leakage, due to potential interception or eavesdropping during voice transmission. Current protections of voice privacy are almost entirely controlled by communication service providers (CSPs), which operate as a black-box to users thus hard to fully trust. To take back the control of user privacy, in this paper, we introduce SecHeadset, an end-to-end solution for secure voice communication based on voice obfuscation, which is plug-and-play and compatible with various CSPs. Our solution involves two parts. First, we design a voice-like noise masking scheme for voice obfuscation. The noise, mimicking voice characteristics, could effectively obscure users' voices while demonstrating resilience against noise reduction methods. Second, we develop a protocol that enables efficient channel state estimation and secure information exchange between two communication entities. Based on this information, we propose a lightweight algorithm for voice retrieval during communication. We develop a prototype of SecHeadset and evaluate its performance with 8 widely-used applications, including Telegram and Skype. It reduces the voice recognition accuracy of various adversaries to below 15% while maintaining communication quality. We also integrate SecHeadset with off-the-shelf portable devices and verify its real-world effectiveness.

Invisible Image Watermarking is crucial for ensuring content provenance and accountability in generative AI. While Gen-AI providers are increasingly integrating invisible watermarking systems, the robustness of these schemes against forgery attacks remains poorly characterized. This is critical, as forging traceable watermarks onto illicit content leads to false attribution, potentially harming the reputation and legal standing of Gen-AI service providers who are not responsible for the content. In this work, we propose WMCopier, an effective watermark forgery attack that operates without requiring any prior knowledge of or access to the target watermarking algorithm. Our approach first models the target watermark distribution using an unconditional diffusion model, and then seamlessly embeds the target watermark into a non-watermarked image via a shallow inversion process. We also incorporate an iterative optimization procedure that refines the reconstructed image to further trade off the fidelity and forgery efficiency. Experimental results demonstrate that WMCopier effectively deceives both open-source and closed-source watermark systems (e.g., Amazon’s system), achieving a significantly higher success rate than existing methods. Additionally, we evaluate the robustness of forged samples and discuss the potential defense against our attack. Code is available at: https://github.com/holdrain/WMCopier.

Semantic caching significantly reduces computational costs and improves efficiency by storing and reusing large language model (LLM) responses. However, existing systems rely primarily on matching individual queries, lacking awareness of multi-turn dialogue contexts, which leads to incorrect cache hits when similar queries appear in different conversational settings. This demonstration introduces ContextCache, a context-aware semantic caching system for multi-turn dialogues. ContextCache employs a two-stage retrieval architecture that first executes vector-based retrieval on the current query to identify potential matches and then integrates current and historical dialogue representations through self-attention mechanisms for precise contextual matching. Evaluation of real-world conversations shows that ContextCache improves precision and recall compared to existing methods. Additionally, cached responses exhibit approximately 10 times lower latency than direct LLM invocation, enabling significant computational cost reductions for LLM conversational applications.

The widespread smart devices raise people’s concerns of being eavesdropped on. To enhance voice privacy, recent studies exploit the nonlinearity in microphone to jam audio recorders with inaudible ultrasound. However, existing solutions solely rely on energetic masking. Their simple-form noise leads to several problems, such as high energy requirements and being easily removed by speech enhancement techniques. Besides, most of these solutions do not support authorized recording, which restricts their usage scenarios. In this paper, we design an efficient yet robust system that can jam microphones while preserving authorized recording. Specifically, we propose a novel phoneme-based noise with the idea of informational masking, which can distract both machines and humans and is resistant to denoising techniques. Besides, we optimize the noise transmission strategy for broader coverage and implement a hardware prototype of our system. Experimental results show that our system can reduce the recognition accuracy of recordings to below 50% under all tested speech recognition systems, which is much better than existing solutions.

Extensive research has revealed that adversarial examples (AE) pose a significant threat to voice-controllable smart devices. Recent studies have proposed black-box adversarial attacks that require only the final transcription from an automatic speech recognition (ASR) system. However, these attacks typically involve many queries to the ASR, resulting in substantial costs. Moreover, AE-based adversarial audio samples are susceptible to ASR updates. In this paper, we identify the root cause of these limitations, namely the inability to construct AE attack samples directly around the decision boundary of deep learning (DL) models. Building on this observation, we propose ALIF, the first black-box adversarial linguistic feature-based attack pipeline. We leverage the reciprocal process of text-to-speech (TTS) and ASR models to generate perturbations in the linguistic embedding space where the decision boundary resides. Based on the ALIF pipeline, we present the ALIF-OTL and ALIF-OTA schemes for launching attacks in both the digital domain and the physical playback environment on four commercial ASRs and voice assistants. Extensive evaluations demonstrate that ALIF-OTL and -OTA significantly improve query efficiency by 97.7% and 73.3%, respectively, while achieving competitive performance compared to existing methods. Notably, ALIF-OTL can generate an attack sample with only one query. Furthermore, our test-of-time experiment validates the robustness of our approach against ASR updates.

Ubiquitous microphones on smart devices considerably raise users’ concerns about speech privacy. Since the microphones are primarily controlled by hardware/software developers, profit-driven organizations can easily collect and analyze individuals’ daily conversations on a large scale with deep learning models, and users have no means to stop such privacy-violating behavior. In this article, we propose UniAP to empower users with the capability of protecting their speech privacy from the large-scale analysis without affecting their routine voice activities. Based on our observation of the recognition model, we utilize adversarial learning to generate quasi-imperceptible perturbations to disturb speech signals captured by nearby microphones, thus obfuscating the recognition results of recordings into meaningless contents. As validated in experiments, our perturbations can protect user privacy regardless of what users speak and when they speak. The jamming performance stability is further improved by training optimization. Additionally, the perturbations are robust against noise removal techniques. Extensive evaluations show that our perturbations achieve successful jamming rates of more than 87% in the digital domain and at least 90% and 70% for common and challenging settings, respectively, in the real-life chatting scenario. Moreover, our perturbations, solely trained on DeepSpeech, exhibit good transferability over other models based on similar architecture.

Inaudible command injection transmits inaudible ultrasounds to inject adversarial speech commands into a voice assistant, therefore manipulating voice control systems (e.g., a garage door or a security camera) for illegitimate purposes. Although the attack is inaudible, we find it does leave visible “footprints”. Such attack “footprints” are the side product due to the interaction between the attack signal (i.e., input) and the acoustic components (i.e., transfer function), so they reflect the hardware characteristics of the sound capture system, including the microphone diaphragm, the low-pass filter, and the analog-to-digital converter. Moreover, unlike the non-linearity distortion that is erasable with signal-shaping techniques, the “footprints” are indelible because they are unrelated to the content of injected commands. We discover two types of indelible “footprints” embedded in the recording spectrogram, namely abnormal interfering noise and abnormal demodulation. A software-based detection method and a portable detector, DolphinTag, are further designed to identify these “footprints”. The software-based method achieves a detection accuracy of 99.8% on the phone models exhibiting abnormal interfering noise, and our DolphinTag achieves 100% detection accuracy which detects the ultrasound attack by actively facilitating the abnormal demodulation.

The proliferation of deepfake content has motivated a surge of detection studies. However, existing detection methods in the audio area exclusively work in English, and there is a lack of data resources in other languages. Cross-lingual deepfake detection, a critical but rarely explored area, urges more study. This paper conducts the first comprehensive study on the cross-lingual perspective of deepfake detection. We observe that English data enriched in deepfake algorithms can teach a detector the knowledge of various spoofing artifacts, contributing to performing detection across language domains. Based on the observation, we first construct a first-of-its-kind cross-lingual evaluation dataset including heterogeneous spoofed speech uttered in the two most widely spoken languages, then explored domain adaptation (DA) techniques to transfer the artifacts detection capability and propose effective and practical DA strategies fitting the cross-lingual scenario. Our adversarial-based DA paradigm teaches the model to learn real/fake knowledge while losing language dependency. Extensive experiments over 137-hour audio clips validate the adapted models can detect fake audio generated by unseen algorithms in the new domain.

Personal voice assistants (PVAs) are increasingly used as interfaces to digital environments. Voice commands are used to interact with phones, smart homes, or cars. In the United States alone, the number of smart speakers, such as Amazon’s Echo and Google Home, has grown by 78% to 118.5 million, and 21% of the U.S. population own at least one device. Given the increasing dependency of society on PVAs, security and privacy of these have become a major concern of users, manufacturers, and policy makers. Consequently, a steep increase in research efforts addressing security and privacy of PVAs can be observed in recent years. While some security and privacy research applicable to the PVA domain predates their recent increase in popularity, many new research strands have emerged. This article provides a survey of the state of the art in PVA security and privacy. The focus of this work is on the security and privacy challenges arising from the use of the acoustic channel. Work that describes both attacks and countermeasures is discussed. We highlight established areas such as voice authentication (VA) and new areas such as acoustic Denial of Service (DoS) that deserve more attention. This survey describes research areas where the threat is relatively well understood but where countermeasures are lacking, for example, in the area of hidden voice commands. We also discuss work that looks at privacy implications; for example, work on management of recording consent. This survey is intended to provide a comprehensive research map for PVA security and privacy.