Research On Data Logic Attack And Detection Of Heavy-duty Industrial Manipulator

Heavy-duty industrial manipulator(HDIM)has an irreplaceable role in the manufacture of heavy equipment,collaborative handling,fine assembly and safety operation of heavy equipment in complex environments in the global industrial field,and has become the core necessary equipment to solve the major demand problems of high-load operation,such as work efficiency,production safety and reducing labor costs.With the rapid development of robotics and Internet of Things technology,the inherent security holes and vulnerabilities of manipulator systems are constantly being exploited,increasing information security risks and the threat of cyber-physical intrusion.Research on the safety of heavy-duty industrial robotic arm systems is relatively inadequate,and damage to the data logic of the robotic arm in network protocols,system data,and physical processes may bring serious security threats.In view of the shortcomings,this dissertation explores potential attacks,and evaluates their attack effects.Then,according to the characteristics of the potential attacks,corresponding detection methods are designed to make up for the deficiencies of existing detection methods.Specifically speaking,we focus on the data logic attack and detection in HDIMs.By analyzing the mechanical structure characteristics,functions and dynamic characteristics of the HDIM,the attack impact analysis model(AIM)was designed.The main contents and contributions of this dissertation are as follows.Then,a data logic attack mechanism is proposed by using the vulnerability of network protocol,system model and physical process.The detection mechanism based on deep learning and dynamic model fusion is studied.Finally,through the hardware-in-the-loop cosimulation and physical system experiments,the effectiveness of the proposed attack model and detection mechanism was verified.(1)A force/position co-loop control algorithm and attack impact analysis model for HDIM is proposed.By analyzing the structural characteristics of HDIM,kinematics and dynamics models are derived.Based on the GNN algorithm to optimize the force control parameters,the GNN-based force/position co-loop control algorithm is established.After that,a cosimulation system of Matlab and Adams was established for deep analysis of the impact of cyber physical attacks on physical quantities such as force and posture.In addition,by modeling factors such as accuracy,integrity,security,and availability based on the co-loop control algorithm,an AIM is proposed to quantify the impact of attacks on HDIMs.(2)A data logic attack method for HDIMs is proposed.Based on the vulnerability of network communication,system model and physical process in data logic,a new type of HDIM data logic attack method is proposed,and the data logic attack model has been established.Then,the hardware device is added to the cosimulation,the hardware-in-the-loop cosimulation model is established,and the typical data logic attack function is modeled and simulated.Finally,according to the AIM,the experimental data is processed and analyzed to obtain the corresponding attack impacts,and the attack impacts in multiple scenarios are quantified and ranked.The effectiveness of the attack model is verified.(3)A detection mechanism based on the fusion of deep learning algorithms and dynamic models is proposed.By optimizing the process of deep belief network data processing and model design,combined with HDIM kinematics and dynamics for modeling,a fusion detection mechanism based on improved deep belief networks(IDBN)and dynamic model is proposed.The performance evaluation test was subsequently carried out.Results show that the classification accuracy of IDBN algorithm was higher than the classification accuracy of the three methods of SVM,BP and DBN,reaching 96.2%.In the dynamic model detection verification,the accuracy reached 94.0%.Experimental results prove that the fusion detection mechanism is effective.Finally,the data logic attack is detected in the master-slave hardware-in-the-loop cosimulation system,and the results show the effectiveness of the proposed mechanism,(4)The cosimulation and physical experiment test platform is designed and implemented.The physical experiment system mainly includes heavy-duty industrial manipulator,Sick displacement sensor,STM32 development board,programmable logic controller(PLC),data encoder,and human-computer interaction(HCI)handle.Subsequently,the composition and development environment of the test system of the HDIM were described,and the basic function realization of the test experiment system was introduced.On this basis,a physical attack and defense experiment was carried out on the data logic attack and the fusion detection mechanism based on the fusion of deep learning and dynamic models,which verified the feasibility of the data logic attack and detection mechanism.