Shi-Feng Sun

Feiyue Xu, Hongsheng Hu, Chaoxiang He et al.

Large Language Models (LLMs) have achieved remarkable success but remain highly susceptible to jailbreak attacks, in which adversarial prompts coerce models into generating harmful, unethical, or policy-violating outputs. Such attacks pose real-world risks, eroding safety, trust, and regulatory compliance in high-stakes applications. Although a variety of attack and defense methods have been proposed, existing evaluation practices are inadequate, often relying on narrow metrics like attack success rate that fail to capture the multidimensional nature of LLM security. In this paper, we present a systematic taxonomy of jailbreak attacks and defenses and introduce Security Cube, a unified, multi-dimensional framework for comprehensive evaluation of these techniques. We provide detailed comparison tables of existing attacks and defenses, highlighting key insights and open challenges across the literature. Leveraging Security Cube, we conduct benchmark studies on 13 representative attacks and 5 defenses, establishing a clear view of the current landscape encompassing jailbreak attacks, defenses, automated judges, and LLM vulnerabilities. Based on these evaluations, we distill critical findings, identify unresolved problems, and outline promising research directions for enhancing LLM robustness against jailbreak attacks. Our analysis aims to pave the way towards more robust, interpretable, and trustworthy LLM systems. Our code is available at Code.

Sheng Hang, Chaoxiang He, Hongsheng Hu et al.

Detecting illicit visual content demands more than image-level NSFW flags; moderators must also know what objects make an image illegal and where those objects occur. We introduce a zero-shot pipeline that simultaneously (i) detects if an image contains harmful content, (ii) identifies each critical element involved, and (iii) localizes those elements with pixel-accurate masks - all in one pass. The system first applies foundation segmentation model (SAM) to generate candidate object masks and refines them into larger independent regions. Each region is scored for malicious relevance by a vision-language model using open-vocabulary prompts; these scores weight a fusion step that produces a consolidated malicious object map. An ensemble across multiple segmenters hardens the pipeline against adaptive attacks that target any single segmentation method. Evaluated on a newly-annotated 790-image dataset spanning drug, sexual, violent and extremist content, our method attains 85.8% element-level recall, 78.1% precision and a 92.1% segment-success rate - exceeding direct zero-shot VLM localization by 27.4% recall at comparable precision. Against PGD adversarial perturbations crafted to break SAM and VLM, our method's precision and recall decreased by no more than 10%, demonstrating high robustness against attacks. The full pipeline processes an image in seconds, plugs seamlessly into existing VLM workflows, and constitutes the first practical tool for fine-grained, explainable malicious-image moderation.

Viet Vo, Xingliang Yuan, Shi-Feng Sun et al.

The security of our data stores is underestimated in current practice, which resulted in many large-scale data breaches. To change the status quo, this paper presents the design of ShieldDB, an encrypted document database. ShieldDB adapts the searchable encryption technique to preserve the search functionality over encrypted documents without having much impact on its scalability. However, merely realising such a theoretical primitive suffers from real-world threats, where a knowledgeable adversary can exploit the leakage (aka access pattern to the database) to break the claimed protection on data confidentiality. To address this challenge in practical deployment, ShieldDB is designed with tailored padding countermeasures. Unlike prior works, we target a more realistic adversarial model, where the database gets updated continuously, and the adversary can monitor it at an (or multiple) arbitrary time interval(s). ShieldDB's padding strategies ensure that the access pattern to the database is obfuscated all the time. Additionally, ShieldDB provides other advanced features, including forward privacy, re-encryption, and flushing, to further improve its security and efficiency. We present a full-fledged implementation of ShieldDB and conduct intensive evaluations on Azure Cloud.

Viet Vo, Shangqi Lai, Xingliang Yuan et al.

Searchable encryption (SE) is one of the key enablers for building encrypted databases. It allows a cloud server to search over encrypted data without decryption. Dynamic SE additionally includes data addition and deletion operations to enrich the functions of encrypted databases. Recent attacks exploiting the leakage in dynamic operations drive rapid development of new SE schemes revealing less information while performing updates; they are also known as forward and backward private SE. Newly added data is no longer linkable to queries issued before, and deleted data is no longer searchable in queries issued later. However, those advanced SE schemes reduce the efficiency of SE, especially in the communication cost between the client and server. In this paper, we resort to the hardware-assisted solution, aka Intel SGX, to ease the above bottleneck. Our key idea is to leverage SGX to take over the most tasks of the client, i.e., tracking keyword states along with data addition and caching deleted data. However, handling large datasets is non-trivial due to the I/O and memory constraints of the SGX enclave. We further develop batch data processing and state compression technique to reduce the communication overhead between the SGX and untrusted server, and minimise the memory footprint in the enclave. We conduct a comprehensive set of evaluations on both synthetic and real-world datasets, which confirm that our designs outperform the prior art.

Shangqi Lai, Xingliang Yuan, Shi-Feng Sun et al.

Network Function Virtualisation (NFV) advances the adoption of composable software middleboxes. Accordingly, cloud data centres become major NFV vendors for enterprise traffic processing. Due to the privacy concern of traffic redirection to the cloud, secure middlebox systems (e.g., BlindBox) draw much attention; they can process encrypted packets against encrypted rules directly. However, most of the existing systems supporting pattern matching based network functions require the enterprise gateway to tokenise packet payloads via sliding windows. Such tokenisation induces a considerable communication overhead, which can be over 100$\times$ to the packet size. To overcome this bottleneck, in this paper, we propose the first bandwidth-efficient encrypted pattern matching protocol for secure middleboxes. We resort to a primitive called symmetric hidden vector encryption (SHVE), and propose a variant of it, aka SHVE+, to achieve constant and moderate communication cost. To speed up, we devise encrypted filters to reduce the number of accesses to SHVE+ during matching highly. We formalise the security of our proposed protocol and conduct comprehensive evaluations over real-world rulesets and traffic dumps. The results show that our design can inspect a packet over 20k rules within 100 $μ$s. Compared to prior work, it brings a saving of $94\%$ in bandwidth consumption.

Cong Zuo, Shi-Feng Sun, Joseph K. Liu et al.

Dynamic searchable symmetric encryption (DSSE) is a useful cryptographic tool in encrypted cloud storage. However, it has been reported that DSSE usually suffers from file-injection attacks and content leak of deleted documents. To mitigate these attacks, forward privacy and backward privacy have been proposed. Nevertheless, the existing forward/backward-private DSSE schemes can only support single keyword queries. To address this problem, in this paper, we propose two DSSE schemes supporting range queries. One is forward-private and supports a large number of documents. The other can achieve backward privacy, while it can only support a limited number of documents. Finally, we also give the security proofs of the proposed DSSE schemes in the random oracle model.

Shangqi Lai, Xingliang Yuan, Shi-Feng Sun et al.

In this paper, we propose GraphSE$^2$, an encrypted graph database for online social network services to address massive data breaches. GraphSE$^2$ preserves the functionality of social search, a key enabler for quality social network services, where social search queries are conducted on a large-scale social graph and meanwhile perform set and computational operations on user-generated contents. To enable efficient privacy-preserving social search, GraphSE$^2$ provides an encrypted structural data model to facilitate parallel and encrypted graph data access. It is also designed to decompose complex social search queries into atomic operations and realise them via interchangeable protocols in a fast and scalable manner. We build GraphSE$^2$ with various queries supported in the Facebook graph search engine and implement a full-fledged prototype. Extensive evaluations on Azure Cloud demonstrate that GraphSE$^2$ is practical for querying a social graph with a million of users.