Recently, many password guessing algorithms have been proposed, seriously threatening cyber security. In this paper, we systematically review over thirty methods for password guessing published between 2016 and 2023. First, we introduce a taxonomy for classifying the existing methods into trawling guessing and targeted guessing. Second, we present an extensive benchmark dataset that can assist researchers and practitioners in successive works. Third, we conduct a bibliometric analysis to present trends in this field and cross-citation between reviewed papers. Further, we discuss the open challenges of password guessing in terms of diverse application scenarios, guessing efficiency, and the combination of traditional and deep learning methods. Finally, this review presents future research directions to guide successive research and development of password guessing.

Efficient exploration in cooperative multi-agent reinforcement learning is still tricky in complex tasks. In this paper, we propose a novel multi-agent collaborative exploration method called Neighborhood Curiosity-based Exploration (NCE), by which agents can explore not only novel states but also new partnerships. Concretely, we use the attention mechanism in graph convolutional networks to perform a weighted summation of features from neighbors. The calculated attention weights can be regarded as an embodiment of the relationship among agents. Then we use the prediction errors of the aggregated features as intrinsic rewards to facilitate exploration. When agents encounter novel states or new partnerships, NCE will produce large prediction errors, resulting in large intrinsic rewards. In addition, agents are more influenced by their neighbors and only interact directly with them in multi-agent systems. Exploring partnerships between agents and their neighbors can enable agents to capture the most important cooperative relations with other agents. Therefore, NCE can effectively promote collaborative exploration even in environments with a large number of agents. Our experimental results show that NCE achieves significant performance improvements on the challenging StarCraft II Micromanagement (SMAC) benchmark.

Shadowsocks(R) is a proxy software based on Socks5, which is the collective name of shadowsocks and shadowsocksR. Shadowsocks(R) is a private protocol without a handshake negotiation mechanism. Peng broke the confidentiality of shadowsocks by exploiting vulnerability in the shadowsocks protocol and decrypted the shadowsocks packets encrypted with none-AEAD encryption options using a redirection attack. Chen et al. started with the cryptographic algorithm used by shadowsocks(R) and preliminarily discussed the confidentiality of user data under the protection of shadowsocks(R) in theory. Based on Chen’s work, this paper further clarifies the shadowsocks(R) protocol format and studies the encryption mechanism of shadowsocks(R) from the perspective of protocol analysis. The vulnerability of the shadowsocks(R) encryption mechanism is found, and an attack method of shadowsocks(R) is proposed. The attack method is a passive attack and can decrypt the shadowsocks packets encrypted with any encryption option. Compared with Peng’s attack method, the method is more effective and more suitable for actual attacks. Finally, some methods to improve the protocol security of shadowsocks(R) are proposed.

The encryption mode of WinRAR3 which does not encrypt the file name uses encryption and compression, the password recovery complexity is high. The existing cracking systems crack on a single CPU or GPU platform. Because the decryption algorithm is slow on the CPU platform, while the decompression algorithm is slow on the GPU platform, the overall performance of the cracking algorithm is not high. This paper studies the mode of CPU and GPU collaborative computing, and proposes an efficient cracking method of encrypted WinRAR3 without encrypting filename. By using the CPU + GPU pipeline cooperation method, the waiting time in the calculation is reduced, and the performance of the algorithm is improved; by using the magic number matching method of compressed files, the decompression calculation can be effectively reduced. The experimental results show that the speed of the cracking algorithm proposed by this paper for 8-digit passwords is 24423/s, which is 2.3 times as fast as before.

MEGA is an end-to-end encrypted cloud storage platform controlled by users. Moreover, the communication between MEGA client and server is carried out under the protection of Transport Layer Security (TLS) encryption, it is difficult to intercept the key data packets in the process of MEGA registration, login, file data upload, and download. These characteristics of MEGA have brought great difficulties to its forensics. This paper presents a method to attack MEGA to provide an effective method for MEGA’s forensics. By debugging the open-source code of MEGA and analyzing the security white paper published, this paper first clarifies the encryption mechanism of MEGA, including the detailed process of registration, login, and file encryption, studies the encryption mechanism of MEGA from the perspective of protocol analysis, and finds out the vulnerability of MEGA encryption mechanism. On this basis, a method to attack MEGA is proposed, and the secret data stored in the MEGA server can be accessed or downloaded; Finally, the efficiency of the attack method is analyzed, and some suggestions to resist this attack method are put forward.

Accurate identification and classification of Darknet traffic is a critical technical challenge for network security supervision. Existing methods predominantly adopt single-modal features and independent classification strategies, making it difficult to effectively handle the hierarchical structural characteristics and complex encryption patterns of Darknet traffic. This paper proposes E2E-MDC (End-to-End Multi-modal Darknet Classification), an end-to-end deep learning framework based on conditional hierarchical mechanism for three-level hierarchical classification of Darknet traffic. The framework integrates four complementary feature extractors—byte-level CNN, packet sequence TCN, bidirectional LSTM, and Transformer—to comprehensively capture traffic patterns from multiple perspectives. A soft conditional hierarchical classification architecture explicitly models dependencies among Level 1 (Darknet type), Level 2 (application category), and Level 3 (specific behavior) by using upper-level prediction probability distributions as conditional input for lower-level classification. On the self-collected Tor dataset containing 8 applications and 8 behavior types, the system achieves 94.90% cascade accuracy, with Level 3 fine-grained classification accuracy reaching 95.02%. On the public Darknet-2020 dataset, cascade accuracy reaches 92.65%, representing improvements of 12% and 15% over existing state-of-the-art methods, respectively, while reducing hierarchical violation rate to below 0.8%. Experimental results demonstrate that the conditional hierarchical mechanism and multi-modal fusion strategy significantly enhance the accuracy and robustness of Darknet traffic classification, providing effective technical support for network security protection.