Research Of Hardware Security For Reconfigurable Computing Processor

With the development of information society,people's demand for information security is increasing,and various security verification methods have been proposed.However,the hardware attacks in recent years have made information security seriously threatened.Hardware attacks can be generated from the inner part of hardware or from the external part.Hardware Trojans are attacks that is generated from the inner part of hardware,which attack hardware by inserting a circuit into hardware to change the hardware function under certain rare conditions.There are many attacks to attack hardware from the external part,such as fault injection attacks and side-channel attacks.In particular,machine learning attacks pose a serious threat to the physical unclonable function in recent years.In order to solve the above problems,this paper proposes two corresponding countermeasures,which make use of the dynamic and partial configuration properties of reconfigurable computing processors.One countermeasure is called Dynamic Resource Management based on Security Value,which is against the hardware trojan attack in reconfigurable computing processors.A metric named security value and a dynamic resource management based on security value are proposed to achieve the highest security under the hardware resource constraints.The simulation results show there is average 10% gain in security value than other countermeasures.The verification on the actual chip also proves the effectiveness of Dynamic Resource Management based on Security Value.The other countermeasure is against the machine learning attacks on physical unclonable function,which is called Physical Unclonable Function based on Router Selection.This countermeasure makes the length and inner connection of PUF with the ability of dynamic configuration,which makes it difficult for adversaries to carry out effective machine learning attacks without affecting the output stability.A corresponding implement strategy of Physical Unclonable Function based on Router Selection is derived based on the consideration of security.Security analysis shows the proposed countermeasure increases the difficulty of attack by 12 orders of magnitude.Further simulation results show that the prediction accuracy of machine learning attacks is less than 53%,which is far lower than the 87% prediction accuracy required for verification.And under natural circumstances it has a prediction accuracy of 50%.Compared with the hardware overhead required for the same security level,the hardware overhead of the proposed countermeasure in this paper is reduced by 5 orders of magnitude,which means the proposed countermeasure can achieve a very high hardware security with very low hardware overhead.The two countermeasures against hardware attacks proposed in this paper both make use of the dynamic and partially configurable characteristics of the reconfigurable computing processor.While reducing overhead,the reconfigurable computing processor is also improved in its ability to resist hardware attacks.