Search

How to correctly acquire the appropriate features is a primary problem in network protocol recognition field. Aiming to avoid the trouble of artificially extracting features in traditional methods and improve recognition accuracy, a network protocol recognition method based on Convolutional Neural Network (CNN) is proposed. The method utilizes deep learning technique, and it processes network flows automatically. Firstly, normalization is performed on the intercepted network flows and they are mapped into two-dimensional matrix which will be used as the input of CNN. Then, an improved classification model named PtrCNN is built, which can automatically extract the appropriate features of network protocols. Finally, the classification model is trained to recognize the network protocols. The proposed approach is compared with several machine learning methods. Experimental results show that the tailored CNN can not only improve protocol recognition accuracy but also ensure the fast convergence of classification model and reduce the classification time.

The second generation of blockchain represented by smart contracts has been developing vigorously in recent years. However, frequent smart contract vulnerability incidents pose a serious risk to blockchain ecosystem security. Since current symbol execution tools often fall into path explosion and thus lead to inefficient detection, this paper expands Mythril's framework to optimize its performance. Firstly, it finds out potential vulnerable code regions using static analysis and identifies critical paths that may have security defects. Then, aiming at the problem that traditional search algorithms cannot actively locate and explore critical paths, this paper presents a multi-objective oriented path search (MOPS) strategy based on path priority. This strategy guides dynamic symbolic execution to cover critical paths quickly, avoiding blind traversal of program execution paths. Finally, it describes security rules and proposes corresponding detection logics for different vulnerability categories. This paper analyzes over 1000 smart contracts extracted from Etherscan. Compared with existing tools based on symbolic execution, the proposed method can reduce time consumption by around 35% while ensuring the accuracy of vulnerability detection. Moreover, existing tools often issue warnings that do not actually cause financial losses. But the proposed method only concentrates on code regions related to transfer of funds, so it can reduce the false alarm rate to some extent.