Conditional Density Estimation (CDE) models the distribution of a target variable given specific inputs and is crucial for applications such as risk assessment, forecasting, and medical diagnosis. Among CDE methods, Deconvolutional Density Networks (DDNs) have shown promise in data-scarce contexts by leveraging local spatial structure. However, they face two primary limitations: (1) overfitting in small datasets due to restricted receptive fields that fail to capture correlations among spatially distant samples, and (2) overly spiky, overconfident predictions arising from reliance on negative log-likelihood (NLL) alone, which neglects global distribution smoothness. In this paper, we propose DDN-Turbo, an enhanced DDN that addresses these challenges by introducing (i) a difference-matrix-based Lipschitz constraint to enforce global smoothness and (ii) a Gaussian Dropout mechanism that injects spatially correlated noise to preserve local structure. Empirical results on multiple benchmarks demonstrate that DDN-Turbo mitigates overfitting, produces more realistic predictive distributions, and outperforms baseline methods in data-scarce scenarios—broadening the applicability of DDNs to real-world problems with limited data.

Face antispoofing detection aims to identify whether the user’s face identity information is legal. Multimodality models generally have high accuracy. However, the existing works of face antispoofing detection have the problem of insufficient research on the safety of the model itself. Therefore, the purpose of this paper is to explore the vulnerability of existing face antispoofing models, especially multimodality models, when resisting various types of attacks. In this paper, we firstly study the resistance ability of multimodality models when they encounter white-box attacks and black-box attacks from the perspective of adversarial examples. Then, we propose a new method that combines mixed adversarial training and differentiable high-frequency suppression modules to effectively improve model safety. Experimental results show that the accuracy of the multimodality face antispoofing model is reduced from over 90% to about 10% when it is attacked by adversarial examples. But, after applying the proposed defence method, the model can still maintain more than 90% accuracy on original examples, and the accuracy of the model can reach more than 80% on attack examples.

The study of software bugs has long been a key area in software security. Dynamic symbolic execution, in exploring the program's execution paths, finds bugs by analyzing all potential dangerous operations. Due to its high coverage and abilities to generate effective testcases, dynamic symbolic execution has attracted wide attention in the research community. However, the success of dynamic symbolic execution is limited due to complex program logic and its difficulty to handle large symbolic data. In our experiments we found that phase-related features of a program often prevents dynamic symbolic execution from exploring deep paths. On the basis of this discovery, we proposed a novel symbolic execution technology guided by program phase characteristics. Compared to KLEE, the most well-known symbolic execution approach, our method is capable of covering more code and discovering more bugs. We designed and implemented pbSE system, which was used to test several commonly used tools and libraries in Linux. Our results showed that pbSE on average covers code twice as much as what KLEE does, and we discovered 21 previously unknown vulnerabilities by using pbSE, out of which 7 are assigned CVE IDs.

With the rapid growth of network bandwidth, increases in CPU cores on a single machine, and application API models demanding more short-lived connections, a scalable TCP stack is performance-critical. Although many clean-state designs have been proposed, production environments still call for a bottom-up parallel TCP stack design that is backward-compatible with existing applications.

Reference counts are widely used in OS kernels for resource management. However, reference counts are not trivial to be used correctly in large scale programs because it is left to developers to make sure that an increment to a reference count is always paired with a decrement. This paper proposes inconsistent path pair checking, a novel technique that can statically discover bugs related to reference counts without knowing how reference counts should be changed in a function. A prototype called RID is implemented and evaluations show that RID can discover more than 80 bugs which were confirmed by the developers in the latest Linux kernel. The results also show that RID tends to reveal bugs caused by developers' misunderstanding on API specifications or error conditions that are not handled properly.

I/O devices are evolving rapidly, while OS optimization is always slower because of its dependence on physical devices. This inevitably prevents latest devices from working with their rating performance, which remains a big problem for performance-critical applications. Though I/O device simulators can help carry out performance evaluation before physical devices are ready, the existing simulator implementations are still unsatisfactory, either having too big overhead or requiring too much extra work. In this paper, we propose kernel built-in device simulation to provide accurate real time evaluations with acceptable extra effort. With the work of simulation well isolated, the overhead is reasonable compared to native environment. A bonding Ethernet interface is implemented in this way and experiments on it confirm the close-to-native performance of the idea.