Aspects of human behavior in cyber security allow more natural security to the user. This research focuses the appearance of anticipating cyber threats and their abstraction hierarchy levels on the mental picture leve...Aspects of human behavior in cyber security allow more natural security to the user. This research focuses the appearance of anticipating cyber threats and their abstraction hierarchy levels on the mental picture levels of human. The study concerns the modeling of the behaviors of mental states of an individual under cyber attacks. The mental state of agents being not observable, we propose a non-stationary hidden Markov chain approach to model the agent mental behaviors. A renewal process based on a nonparametric estimation is also considered to investigate the spending time in a given mental state. In these approaches, the effects of the complexity of the cyber attacks are taken into account in the models.展开更多
在自然语言处理(Natural Language Processing,NLP)领域,后门攻击已成为现代NLP应用的重大威胁,严重影响系统的安全性与可靠性。尽管文本领域已提出多种防御策略,但在不接触中毒数据集也不参与后门训练过程时,面对复杂的攻击场景,现有...在自然语言处理(Natural Language Processing,NLP)领域,后门攻击已成为现代NLP应用的重大威胁,严重影响系统的安全性与可靠性。尽管文本领域已提出多种防御策略,但在不接触中毒数据集也不参与后门训练过程时,面对复杂的攻击场景,现有方法仍难以有效应对。为此,提出一种基于机器遗忘的文本后门攻击防御方法NLPShield。该方法仅需少量干净样本,通过基于错误标注的训练和干净神经元剪枝两个关键阶段,实现对文本后门攻击的有效防御。实验在SST-2和AGNews数据集上进行,结果显示,在保持较高干净准确率的情况下,NLPShield方法相较于现有最先进基线防御方法,平均能将攻击成功率降低24.83%。这表明NLPShield方法能显著提升多种后门攻击的防御效果,切实有效地缓解文本后门攻击。展开更多
随着电力配变网络基础设施规模的不断扩大,各类安全二次设备、边缘终端节点和业务系统产生的信息通信流量数据在格式、协议、语义特征层面存在显著差异。主要存在现有缓解框架缺乏多源异构网络异常流量检测数据归一化处理算法,网络攻击...随着电力配变网络基础设施规模的不断扩大,各类安全二次设备、边缘终端节点和业务系统产生的信息通信流量数据在格式、协议、语义特征层面存在显著差异。主要存在现有缓解框架缺乏多源异构网络异常流量检测数据归一化处理算法,网络攻击行为分析依赖人工特征提取的规则引擎,以及难以确定有效的网络攻击缓解措施等痛点。针对以上痛点,提出了一种基于归一化处理和TrafficLLM的网络攻击缓解框架(Network Attack Mitigation Framework Based on Normalized Processing and TrafficLLM,NAMF-NPTLLM)。该框架涵盖数据解析、归一化处理、模型微调和生成攻击缓解方案4个核心阶段。首先,在特征选择阶段,通过构建集成学习模型,融合多类基学习器的特征评估结果,精准提取对分类结果影响较大的关键特征。其次,将选取的关键特征通过归一化处理,生成统一的自然语言token序列形式表达,为该网络攻击缓解框架的流量异常分析TrafficLLM模型提供标准化输入。然后,对TrafficLLM模型进行微调,使该模型能够理解提示词模板指令并学习攻击行为的流量模式。最后,通过微调后的大模型进行推理,生成攻击缓解指令,使得该框架能够根据攻击行为特征动态调整网络攻击缓解策略。通过在CIC-DDoS2019数据集上进行实验验证,与传统方法相比,该框架将网络攻击行为分类的准确率达到99.80%,提高了1.3%。实验结果表明,该框架对于缓解海量多源异构电力网络终端流量攻击,具有更好的准确性和有效性。展开更多
The available studies in the literature on physical and mathematical modeling of the argon oxygen decarburization (AOD) process of stainless steel have briefly been reviewed. The latest advances made by the author wi...The available studies in the literature on physical and mathematical modeling of the argon oxygen decarburization (AOD) process of stainless steel have briefly been reviewed. The latest advances made by the author with his research group have been summarized. Water modeling was used to investigate the fluid flow and mixing characteristics in the bath of an 18 t AOD vessel, as well as the 'back attack' action of gas jets and its effects on the erosion and wear of the refractory lining, with sufficiently full kinematic similarity. The non rotating and rotating gas jets blown through two annular tuyeres, respectively of straight tube and spiral flat tube type, were employed in the experiments. The geometric similarity ratio between the model and its prototype (including the straight tube type tuyeres) was 1:3. The influences of the gas flow rate, the angle included between the two tuyeres and other operating parameters, and the suitability of the spiral tuyere as a practical application, were examined. These latest studies have clearly and successfully brought to light the fluid flow and mixing characteristics in the bath and the overall features of the back attack phenomena of gas jets during the blowing, and have offered a better understanding of the refining process. Besides, mathematical modeling for the refining process of stainless steel was carried out and a new mathematical model of the process was proposed and developed. The model performs the rate calculations of the refining and the mass and heat balances of the system. Also, the effects of the operating factors, including adding the slag materials, crop ends, and scrap, and alloy agents; the non isothermal conditions; the changes in the amounts of metal and slag during the refining; and other factors were all considered. The model was used to deal with and analyze the austenitic stainless steel making (including ultra low carbon steel) and was tested on data of 32 heats obtained in producing 304 grade steel in an 18 t AOD vessel. The changes in the bath composition and temperature during the refining process with time can be accurately predicted using this model. The model can provide some very useful information and a reliable basis for optimizing the process practice of the refining of stainless steel and control of the process in real time and online.展开更多
文摘Aspects of human behavior in cyber security allow more natural security to the user. This research focuses the appearance of anticipating cyber threats and their abstraction hierarchy levels on the mental picture levels of human. The study concerns the modeling of the behaviors of mental states of an individual under cyber attacks. The mental state of agents being not observable, we propose a non-stationary hidden Markov chain approach to model the agent mental behaviors. A renewal process based on a nonparametric estimation is also considered to investigate the spending time in a given mental state. In these approaches, the effects of the complexity of the cyber attacks are taken into account in the models.
文摘在自然语言处理(Natural Language Processing,NLP)领域,后门攻击已成为现代NLP应用的重大威胁,严重影响系统的安全性与可靠性。尽管文本领域已提出多种防御策略,但在不接触中毒数据集也不参与后门训练过程时,面对复杂的攻击场景,现有方法仍难以有效应对。为此,提出一种基于机器遗忘的文本后门攻击防御方法NLPShield。该方法仅需少量干净样本,通过基于错误标注的训练和干净神经元剪枝两个关键阶段,实现对文本后门攻击的有效防御。实验在SST-2和AGNews数据集上进行,结果显示,在保持较高干净准确率的情况下,NLPShield方法相较于现有最先进基线防御方法,平均能将攻击成功率降低24.83%。这表明NLPShield方法能显著提升多种后门攻击的防御效果,切实有效地缓解文本后门攻击。
文摘随着电力配变网络基础设施规模的不断扩大,各类安全二次设备、边缘终端节点和业务系统产生的信息通信流量数据在格式、协议、语义特征层面存在显著差异。主要存在现有缓解框架缺乏多源异构网络异常流量检测数据归一化处理算法,网络攻击行为分析依赖人工特征提取的规则引擎,以及难以确定有效的网络攻击缓解措施等痛点。针对以上痛点,提出了一种基于归一化处理和TrafficLLM的网络攻击缓解框架(Network Attack Mitigation Framework Based on Normalized Processing and TrafficLLM,NAMF-NPTLLM)。该框架涵盖数据解析、归一化处理、模型微调和生成攻击缓解方案4个核心阶段。首先,在特征选择阶段,通过构建集成学习模型,融合多类基学习器的特征评估结果,精准提取对分类结果影响较大的关键特征。其次,将选取的关键特征通过归一化处理,生成统一的自然语言token序列形式表达,为该网络攻击缓解框架的流量异常分析TrafficLLM模型提供标准化输入。然后,对TrafficLLM模型进行微调,使该模型能够理解提示词模板指令并学习攻击行为的流量模式。最后,通过微调后的大模型进行推理,生成攻击缓解指令,使得该框架能够根据攻击行为特征动态调整网络攻击缓解策略。通过在CIC-DDoS2019数据集上进行实验验证,与传统方法相比,该框架将网络攻击行为分类的准确率达到99.80%,提高了1.3%。实验结果表明,该框架对于缓解海量多源异构电力网络终端流量攻击,具有更好的准确性和有效性。
文摘The available studies in the literature on physical and mathematical modeling of the argon oxygen decarburization (AOD) process of stainless steel have briefly been reviewed. The latest advances made by the author with his research group have been summarized. Water modeling was used to investigate the fluid flow and mixing characteristics in the bath of an 18 t AOD vessel, as well as the 'back attack' action of gas jets and its effects on the erosion and wear of the refractory lining, with sufficiently full kinematic similarity. The non rotating and rotating gas jets blown through two annular tuyeres, respectively of straight tube and spiral flat tube type, were employed in the experiments. The geometric similarity ratio between the model and its prototype (including the straight tube type tuyeres) was 1:3. The influences of the gas flow rate, the angle included between the two tuyeres and other operating parameters, and the suitability of the spiral tuyere as a practical application, were examined. These latest studies have clearly and successfully brought to light the fluid flow and mixing characteristics in the bath and the overall features of the back attack phenomena of gas jets during the blowing, and have offered a better understanding of the refining process. Besides, mathematical modeling for the refining process of stainless steel was carried out and a new mathematical model of the process was proposed and developed. The model performs the rate calculations of the refining and the mass and heat balances of the system. Also, the effects of the operating factors, including adding the slag materials, crop ends, and scrap, and alloy agents; the non isothermal conditions; the changes in the amounts of metal and slag during the refining; and other factors were all considered. The model was used to deal with and analyze the austenitic stainless steel making (including ultra low carbon steel) and was tested on data of 32 heats obtained in producing 304 grade steel in an 18 t AOD vessel. The changes in the bath composition and temperature during the refining process with time can be accurately predicted using this model. The model can provide some very useful information and a reliable basis for optimizing the process practice of the refining of stainless steel and control of the process in real time and online.
