Security Analysis Of Image Encryption Schemes Based On Chaos

The rapid development and extensive application of scientific technology,especially network technology,brings unprecedented convenience to people's daily life,but what follows is the endless interference caused by information leakage.As an intuitive and visual way of message expression,image is widely used in network communication.Therefore,it has become an important topic to resist the leakage of image information.Image encryption technology is the most important method to protect image information.Chaotic systems are widely concerned because of their own characteristics,and are very suitable for image encryption.In recent years,researchers have proposed a variety of image encryption techniques based on chaos with different methods.But these encryption techniques do not always fight against existing types of attacks.Therefore,it is significant to analyze the security of image encryption algorithm based on chaos.This thesis analyzes the advantages and disadvantages of three recently proposed chaos-based image encryption models,and uses the statistical attack or the chosen plaintext attack to successfully break the three models.In view of the defects in each model,their corresponding improved algorithms are proposed to resist the chosen plaintext attack.In the experimental part,the author uses the existing security indexes to evaluate the improved models.The result shows that the three improved algorithms can resist exhaustive attack,statistical attack,noise attack and chosen plaintext attack very well.The main research results of this thesis are as follows:(1)Analyse the security of an image encryption model based on three dimensional bit matrix permutation(TDBMP).This thesis finds the defects of the bit matrix scrambling in the encrypted model,that is,the parameter space of the Arnold map is small,and the first location of the map is always mapped to itself.In view of these defects of Arnold map,method to enlarge parameter space and hide fixed point is designed.Because that TDBMP is an asynchronous permutation-diffusion algorithm,the attacker can obtain the equivalent keys of permutation process and diffusion process,respectively,and then break the whole model.(2)Analyse the security of a synchronous permutation-diffusion image encryption technology(SPDT).The thesis studies SPDT model in detail,and finds that it is essentially asynchronous,and the rules generated by the DNA method used in the diffusion stage are not homogeneous.Therefore,the SPDT is easy to be statistically attacked and the corresponding statistical attack scheme is designed in the thesis.(3)Analyse the security of a color image encryption model(APDS)based on one dimension chaotic map.The model is an asynchronous permutation-diffusion mechanism,and the equivalent keys of the model cannot change with the difference of the plaintext,so it is easy to attacked by chosen plaintext attack.In this thesis,a chosen plaintext attack scheme is designed to break the APDS.(4)Two methods to resist the chosen plaintext attack are put forward.The common defects of the above three encryption schemes are that they all have the fixed equivalent keys and are based on asynchronous permutation-diffusion mechanism.The fixed equivalent keys,namely,when the secret keys of the encryption system are unchanged,the equivalent keys are fixed values regardless of the change in the plain image,which allow the attacker to guess the fixed equivalent keys by comparing different pairs of plain and cipher images.The asynchronous permutation-diffusion mechanism can enable the attacker to obtain the equivalent keys in the process of permutation and diffusion respectively,and break the whole model.The thesis analyses and finds that if the equivalent keys of encryption model are not fixed or the model is synchronous permutation-diffusion,it will be difficult to break by the chosen plaintext attack.Therefore,the plaintext eigenvalue method and the synchronous permutation-diffusion mechanism are designed to resist the chosen plaintext attack.(5)The improved algorithms of three models are proposed to resist the chosen plaintext attack.For the TDBMP and SPDT models,the plaintext eigenvalue method is used to make the equivalent keys of the encryption system differ with the input of the plain image.In view of the APDS model,this thesis changes some steps in the original model,and designs a synchronous permutation-diffusion mechanism.(6)Analyse the security of the three improved models from four aspects: exhaustive attack,statistical attack,chosen plaintext attack and clipping attack,and analyse the time complexity of the improved model in the evaluation phase.