With the emergence of increasingly growing network threats, network security becomes a major issue which causes huge existing and potential losses, such as WannaCry. Various methods had been adopted to maintain network security, among which Intrusion Detection System (IDS) is one of the most essential parts of cybersecurity to defense against sophisticated and ever-growing network attacks. A number of researchers have studied comprehensive datasets and effective approaches to build IDS. Machine learning models are also applied in IDS and obtained considerable results in building better network security system. In this paper, we proposed a two-stage IDS based on machine learning models RF and SVM optimized with Feature Selection algorithm CFS. We also conducted experiments on NSL-KDD benchmark datasets to evaluate the performance of the two-stage IDS, comparing against RF and SVM models respectively. The results demonstrated that our proposed two-stage IDS outperformed RF and SVM with an increase from 4.31% to 14.62% in Precision and a reduction of 93.84% in time than SVM.

Stance detection, the task of identifying the stance expressed in a text toward a specific target, is essential for analyzing public opinion across diverse domains. The existing approaches primarily focus on modeling the semantic relationship between the text and target, but they often struggle when the target is implicit or indirectly referenced. In real-world scenarios, stance is frequently conveyed through references to related entities, events, or contextual implications, making stance detection particularly challenging. To tackle this challenge, we propose a novel framework that leverages large language models to construct a heterogeneous stance network from textual data. Based on this network, we develop two complementary methodologies tailored for distinct application scenarios: (1) In a supervised setting, we employ a graph neural network approach to learn stance representations from the heterogeneous stance network, enhancing stance prediction performance. (2) For zero-shot stance detection, we introduce an LLM-based method that leverages the heterogeneous stance network to infer stance without task-specific supervision. The experimental results on benchmark datasets demonstrate that our methods outperform the existing approaches, highlighting their effectiveness in both supervised and zero-shot scenarios.

To describe the dynamic property of trust relationship, we propose a time-related trust model and extend Jøsang's subjective logic to fit for time-related trust model. The extension includes propositional conjunction, disjunction and negation for traditional logic and discounting and consensus operators that are evidential operators specially designed for the propagation and computation of trust relationships. With the extension of subjective logic for time-related trust, our time-related trust model is suitable to model the dynamic trust relationship in practice. Finally an example of reputation assessment is offered to demonstrate the usage of our trust model.

Biometric authentication is up-and-coming to replace the traditional identity authentication method (e.g., passwords, PIN, identification cards) for its convenience and intelligence. With more and more users using this method, the database becomes more extensive, and the functions are seriously challenged. Data outsourcing has advantages in terms of convenience and cost savings, so it has attracted much research effort. However, due to the biometric's immutability of the whole life, it is extremely sensitive, and disclosing it to a third party is undesirable. In this paper, we address the issue of securely outsourcing biometric database. We propose a novel boolean searchable symmetric encryption (SSE) to construct a secure interactive protocol when outsourcing. A new encrypted database construction method was proposed, using the more efficient boolean vectors. Based on this, We suggest three kinds of expressive SSE, supporting disjunctive query, boolean query, and lightweight settings. We prove the schemes' correctness and security theoretically. Our constructions use simple cryptographic tools, such as symmetric cryptography and pseudo-random functions. They are straightforward to understand and easy to implement. The experiments show that all our schemes are practical and more efficient than the existing methods.

Instant messaging (IM) has become one of the most popular online communication tools among consumer and enterprise IM users. It provides instant message delivery, as well as convenient file transfer services. The increasing popularity and functionalities of IM programs have made it increasingly attractive for attackers, especially for worm writers. IM contact list offers worm an easy way of finding potential victims so that the worm could achieve a surprising spreading speed. This paper first presents our experimental results of simulating IM worm propagation in the logical network defined by IM contact lists, which is reported to be a scale-free network. Then, the existing proposals for detecting and containing IM worm epidemics are discussed. At last, a new algorithm for this purpose is presented, which is based on the observation of the bi-directional nature of IM worm traffic, and its advantages and possible improvements in implementation are analyzed. The simulation results show the proposed algorithm is of significant effect on restricting IM worm propagation.

Digital microfluidic biochips (DMFBs) play an important role in the biochemical industry due to their low cost, portability and high efficiency. It's reported that the biochip growth is twice as large as before in the market. However, the resolution of microfluidic biochip security has not kept pace with the development of business opportunities, and there are still many major problems. One of the problems is the hardware Trojans. According to the classification of hardware Trojans based on biochips, in general-purpose DMFBs, DMFB is configured to execute specific biochemical protocols. Attackers can modify these protocols by changing the actuation sequences in the field. In this paper, due to the cyberphysical microfluidic biochips, we consider using an efficient protect model, based on Hamming distance, to detect differences between two sequences. This protect model can be used to find out if the actuation sequence has been tampered with. It is simple, effective and it has a low error rate. The experimental results prove the feasibility of our proposed method.