Verifiable data streaming (VDS) enables the client to incrementally store a sequence of ordered data on an untrusted cloud server, and verify the validity of the retrieved data. Moreover, the client can replace a data with another value. The common security problem caused by updating operation is the cloud server may use old authentication information to make expired data pass the verification. To solve this problem, the known approaches use the cryptographic accumulator that actually influences the performance of VDS scheme. The main concerns can be generalized as how to design a VDS scheme without cryptographic accumulator, in such a way that further optimizes the performance of VDS scheme. We put forward the idea to convert the standard digital signature relevant to the updated data into chameleon digital signature whose non-transferability is the key to solve the problem. This is the first attempt to securely authenticate the dynamic data without cryptographic accumulator. In the proposed VDS scheme, the client's local storage overhead, computation overheads of the cloud server in responding to a query and updating the data are constant. As the experimental results shown, the proposed VDS scheme outperforms the scheme in terms of the efficiency.

Extracting causal relationships in medical texts is essential for improving clinical decision support systems and constructing comprehensive medical knowledge graphs. This paper presents a novel model for extracting causal, conditional, and hypernym relationships from Chinese medical texts, combining DeBERTa's advanced contextual encoding with GPLinker's efficient entity and relationship extraction mechanisms. Our approach computes the score matrix of entities and their causal relationships, followed by a decoding process to obtain the final predictions. On the CMedCausal dataset, comparative experiments highlight our model's superior performance in terms of precision, recall, and F1 score, demonstrating its robustness and effectiveness in managing overlapping and nested entities and accurately extracting causal relationships in Chinese medical texts.

Cloud computing has become prevalent due to its extensive storage resources and robust computational capacities. To protect data security and privacy, data owners opt for uploading encrypted data to the cloud. Flexible sharing of these encrypted data in a group of users is a critical functionality in cloud storage. In addition, given that users may exit the group, revocation becomes a crucial requirement in group data-sharing systems. The Key-Aggregate Cryptosystem (KAC) has become a promising mechanism for group data sharing. The decryption rights for any set of ciphertexts can be efficiently delegated by distributing a constant-size aggregate key, while the confidentiality of other ciphertexts outside the set is maintained. However, in previous KAC schemes, revocation remains a challenging task regarding key update, ciphertext re-encryption, and collision resistance. In this paper, we propose a Key-Aggregate Cryptosystem with User Revocation (KAC-UR) scheme to overcome this challenge. The KAC-UR scheme not only achieves flexible data sharing, but also can perform secure and efficient user revocation with properties including collision resistance, revocation without data owner-user communication, and constant ciphertext size. The KAC-UR scheme also enables the cloud server to perform partial decryption, thereby significantly alleviating the computational burden for users. The KAC-UR scheme is chosen plaintext attack secure under the decisional Bilinear Diffie-Hellman Exponent assumption.

Abstract Multi-keyword ranked searchable encryption (MRSE) supports multi-keyword contained in one query and returns the top-k search results related to the query keyword set. It realized effective search on encrypted data. Most previous works about MRSE can only make the complete keyword search and rank on the server-side. However, with more practice, users may not be able to express some keywords completely when searching. Server-side ranking increases the possibilities of the server inferring some keywords queried, leading to the leakage of the user’s sensitive information. In this paper, we propose a new MRSE system named ‘multi-keyword ranked searchable encryption with the wildcard keyword (MRSW)’. It allows the query keyword set to contain a wildcard keyword by using Bloom filter (BF). Using hierarchical clustering algorithm, a clustering Bloom filter tree (CBF-Tree) is constructed, which improves the efficiency of wildcard search. By constructing a modified inverted index (MII) table on the basis of the term frequency-inverse document frequency (TF-IDF) rule, the ranking function of MRSW is performed by the user. MRSW is proved secure under adaptive chosen-keyword attack (CKA2) model, and experiments on a real data set from the web of science indicate that MRSW is efficient and practical.

Key-aggregate searchable encryption (KASE) schemes support selective data sharing and keyword-based ciphertext searching by using the constant-size shared key and trapdoor, making these schemes attractive for resource-constrained users to store, share, and search encrypted data in public clouds. However, most previously proposed KASE schemes suffer from our proposed “off-line keyword guessing attack (KGA)” and some other weaknesses. Consequently, they fail to gain the keyword ciphertext indistinguishability and trapdoor indistinguishability, which are vital security goals of searchable encryption. Inspired by the relationship of public key encryption with keyword search (PEKS) and KASE, we design a new KASE scheme called key-aggregate searchable encryption with a designated server (dKASE). The dKASE scheme achieves our proposed keyword ciphertext indistinguishability against chosen keyword attack (KC-IND-CKA) and keyword trapdoor indistinguishability against keyword guessing attack (KT-IND-KGA) security models, where the latter model captures off-line KGA. Then, we extend the dKASE scheme to verifiable dKASE in multi-owner setting (dVKASEM) scheme. With dVKASEM, when multiple data owners authorize a user to access data, the user merely needs to store his single key and generate a single trapdoor to query these owners’ data. Besides, the adoption of the aggregate signature significantly reduces the overhead of verifying whether data has been tampered with. Performance analysis illustrates that our schemes are efficient.

Generalized zero-shot learning (GZSL) aims to recognize any test instance coming either from a known class or from a novel class that has no training instance. To synthesize training instances for novel classes and thus resolving GZSL as a common classification problem, we propose a Dissimilarity Representation Learning (DSS) method. Dissimilarity representation is to represent a specific instance in terms of its (dis)similarity to other instances in a visual or attribute based feature space. In the dissimilarity space, instances of the novel classes are synthesized by an end-to-end optimized neural network. The neural network realizes two-level feature mappings and domain adaptions in the dissimilarity space and the attribute based feature space. Experimental results on five benchmark datasets, i.e., AWA, AWA$_2$, SUN, CUB, and aPY, show that the proposed method improves the state-of-the-art with a large margin, approximately 10% gain in terms of the harmonic mean of the top-1 accuracy. Consequently, this paper establishes a new baseline for GZSL.

Zero-shot learning (ZSL) which aims to recognize unseen classes with no labeled training sample, efficiently tackles the problem of missing labeled data in image retrieval. Nowadays there are mainly two types of popular methods for ZSL to recognize images of unseen classes: probabilistic reasoning and feature projection. Different from these existing types of methods, we propose a new method: sample construction to deal with the problem of ZSL. Our proposed method, called Imagination Based Sample Construction (IBSC), innovatively constructs image samples of target classes in feature space by mimicking human associative cognition process. Based on an association between attribute and feature, target samples are constructed from different parts of various samples. Furthermore, dissimilarity representation is employed to select high-quality constructed samples which are used as labeled data to train a specific classifier for those unseen classes. In this way, zero-shot learning is turned into a supervised learning problem. As far as we know, it is the first work to construct samples for ZSL thus, our work is viewed as a baseline for future sample construction methods. Experiments on four benchmark datasets show the superiority of our proposed method.