摘要
由于不同尺度上的通信软件数据不具有可比性,导致在对DDoS攻击进行检测时,可靠性偏低,为此,提出基于深度学习的通信软件DDoS攻击检测方法研究。将所有的通信软件数据转换为0到1之间的值后,通过计算出所有可能的灰度级组合在单位距离和特定方向上的出现频率构建了通信软件数据灰度共生矩阵,将通信软件数据灰度共生矩阵的均值和方差特征作为基准,利用SHA-256哈希函数将数据转换为二进制编码形式。将重构后的数据输入到包含交叉熵损失函数的CNN网络中,根据更新偏置与通信软件数据灰度共生矩阵特征参数之间的关系,确定DDoS攻击数据。在测试结果中,PPV始终稳定在0.88以上,TPR稳定在0.92以上,具有较高的可靠性。
Because the communication software data on different scales is not comparable,the re-liability is low when detecting DDoS attacks.Therefore,the detection method of DDoS attacks of communication software based on decp learning is proposed.After converting all the commu-nication software data into valuces betwecn O and 1,the symbiosis matrix of the communication software data is constructed by calculating the frequency of occurrence per unit distance and spe-cific direction,using the mean and variance characteristics of the communication software data as the benchmark,and the data is converted into binary coding form using SHA-256 hash function.The reconstructed data was fed into the CNN network containing the cross-entropy loss func-tion,and the DDoS attack data was determined based on the relationship between the update bias and the characteristic parameters of the gray-scale symbiosis matrix of the communication soft-ware data.In the test results,PPV was always stable above O.88,and TPR was stable above 0.92,with high reliability.
作者
赵菊芳
ZHAO Jufang(Guangzhou College of Commerce,Guangdong guangzhou 513363,China)
出处
《长江信息通信》
2024年第6期102-104,共3页
Changjiang Information & Communications
