Your Fixed Watermark is Fragile: Towards Semantic-Aware Watermark for EaaS Copyright Protection

Embedding-as-a-Service (EaaS) has emerged as a successful business pattern but faces significant challenges related to various forms of copyright infringement, including API misuse and different attacks. Various studies have proposed backdoor-based watermarking schemes to protect the copyright of EaaS services. In this paper, we reveal that previous watermarking schemes possess semantic-independent characteristics and propose the Semantic Perturbation Attack (SPA). Our theoretical and experimental analyses demonstrate that this semantic-independent nature makes current watermarking schemes vulnerable to adaptive attacks that exploit semantic perturbations test to bypass watermark verification. To address this vulnerability, we propose the Semantic Aware Watermarking (SAW) scheme, a robust defense mechanism designed to resist SPA, by injecting a watermark that adapts to the text semantics. Extensive experimental results across multiple datasets demonstrate that the True Positive Rate (TPR) for detecting watermarked samples under SPA can reach up to more than 95%, rendering previous watermarks ineffective. Meanwhile, our watermarking scheme can resist such attack while ensuring the watermark verification capability. Our code is available at https://github.com/Zk4-ps/EaaS-Embedding-Watermark.

Prompt-Guided Internal States for Hallucination Detection of Large Language Models

Large Language Models (LLMs) have demonstrated remarkable capabilities across a variety of tasks in different domains. However, they sometimes generate responses that are logically coherent but factually incorrect or misleading, which is known as LLM hallucinations. Data-driven supervised methods train hallucination detectors by leveraging the internal states of LLMs, but detectors trained on specific domains often struggle to generalize well to other domains. In this paper, we aim to enhance the cross-domain performance of supervised detectors with only in-domain data. We propose a novel framework, prompt-guided internal states for hallucination detection of LLMs, namely PRISM. By utilizing appropriate prompts to guide changes in the structure related to text truthfulness within the LLM's internal states, we make this structure more salient and consistent across texts from different domains. We integrated our framework with existing hallucination detection methods and conducted experiments on datasets from different domains. The experimental results indicate that our framework significantly enhances the cross-domain generalization of existing hallucination detection methods.

BadActs: A Universal Backdoor Defense in the Activation Space

Backdoor attacks pose an increasingly severe security threat to Deep Neural Networks (DNNs) during their development stage. In response, backdoor sample purification has emerged as a promising defense mechanism, aiming to eliminate backdoor triggers while preserving the integrity of the clean content in the samples. However, existing approaches have been predominantly focused on the word space, which are ineffective against feature-space triggers and significantly impair performance on clean data. To address this, we introduce a universal backdoor defense that purifies backdoor samples in the activation space by drawing abnormal activations towards optimized minimum clean activation distribution intervals. The advantages of our approach are twofold: (1) By operating in the activation space, our method captures from surface-level information like words to higher-level semantic concepts such as syntax, thus counteracting diverse triggers; (2) the fine-grained continuous nature of the activation space allows for more precise preservation of clean content while removing triggers. Furthermore, we propose a detection module based on statistical information of abnormal activations, to achieve a better trade-off between clean accuracy and defending performance.

Semantic-Preserving Linguistic Steganography by Pivot Translation and Semantic-Aware Bins Coding

Linguistic steganography (LS) aims to embed secret information into a highly encoded text for covert communication. It can be roughly divided to two main categories, i.e., modification based LS (MLS) and generation based LS (GLS). Unlike MLS that hides secret data by slightly modifying a given text without impairing the meaning of the text, GLS uses a trained language model to directly generate a text carrying secret data. A common disadvantage for MLS methods is that the embedding payload is very low, whose return is well preserving the semantic quality of the text. In contrast, GLS allows the data hider to embed a high payload, which has to pay the high price of uncontrollable semantics. In this paper, we propose a novel LS method to modify a given text by pivoting it between two different languages and embed secret data by applying a GLS-like information encoding strategy. Our purpose is to alter the expression of the given text, enabling a high payload to be embedded while keeping the semantic information unchanged. Experimental results have shown that the proposed work not only achieves a high embedding payload, but also shows superior performance in maintaining the semantic consistency and resisting linguistic steganalysis.

Exploiting Language Model for Efficient Linguistic Steganalysis: An Empirical Study

Recent advances in linguistic steganalysis have successively applied CNNs, RNNs, GNNs and other deep learning models for detecting secret information in generative texts. These methods tend to seek stronger feature extractors to achieve higher steganalysis effects. However, we have found through experiments that there actually exists significant difference between automatically generated steganographic texts and carrier texts in terms of the conditional probability distribution of individual words. Such kind of statistical difference can be naturally captured by the language model used for generating steganographic texts, which drives us to give the classifier a priori knowledge of the language model to enhance the steganalysis ability. To this end, we present two methods to efficient linguistic steganalysis in this paper. One is to pre-train a language model based on RNN, and the other is to pre-train a sequence autoencoder. Experimental results show that the two methods have different degrees of performance improvement when compared to the randomly initialized RNN classifier, and the convergence speed is significantly accelerated. Moreover, our methods have achieved the best detection results.