Mobile Code Protection Against Malicious Hosts

Recently, mobile code technology has a great deal of development and promises well. However the security problem with mobile code technology obstructs its wide application in real business. Among these security problems, how to protect mobile code from malicious host is a new issue that cannot be dealt by traditional technologies. So protection technologies of mobile code in malicious host attract much attention and there are very important theoretical and practical significance.Confidentiality is the most essential and difficult that remains open among mobile code protection problems. This paper treated Java byte code as target and deeply researched all kind of technologies about confidentiality of mobile code. For Intellectual Property protection and anti software pirate, the paper also deeply researched software watermarking technology.A mobile agent protection scheme based on non-interactive secure function evaluation was presented using an optimized oblivious transfer (OT) protocol. Most overheads of an OT protocol are modular exponentiations that are computationally intensive tasks. By expanding from one-out-of-two OT protocol to one-out-N OT protocol and combining N one-out-of- two OT protocol to one one-out-N OT protocol, define a new simultaneous one-out-N OT protocol, and the number of modular exponentiation is consequently reduced to small constant. Through combining the simultaneous OT protocol and the encrypted circuit construction, obtain a non-interactive secure function evaluation protocol. Source host proceeds as Alice, and multiple hosts proceed as Bob. Each host contributes the part of encrypted circuit representing its function, and thus the resulting encrypted circuit is a cascade of sub-circuits. It is proved that the new scheme has lower overhead, and is more security and easier to implement than other similar systems.We also analyzed Java virtual machine model and class file format and presented a formal model of Java byte code obfuscation. We demonstrated some interesting properties of obfuscating transformation and proved that for Java byte code it is possible to create a single-pass algorithm of obfuscation. We described all kinds of obfuscating methods as viewed from byte code and designed effective method to produce redundant code. We also discussed all kinds of methods of quality measurements to estimate efficiency of the obfuscation process. On the base of theoretical analysis we proposed a general obfuscation algorithm of byte code that is equally efficient but simpler thanalgorithms making full analysis of programs written in high-level languages, and it also has more advantages in flexibility and scalability.In the end, a novel blind software watermarking scheme was proposed based on statistic distribution of instructions. It is one of the most robust software watermarking schemes. The main idea is that randomly select two disjoint subsets of equal size out of instruction set and this selection is the watermark which is superimposed by changing the elements of one subset by adding a positive integer factor. For detecting mark, compute the means of two subsets. If the difference of two means exceeded the threshold, the mark is found. This blind scheme is very robust and is immune to most known attacks because the mark is spread over the whole instructions distribution. Its effectiveness and robustness is derived from hypotheses test theory. It is implemented treating Java byte code as targets and is validated by a lot of experiments.We deeply researched protection technologies of mobile code in malicious host, and widely explored non-interactive secure function evaluation, code obfuscation, and software watermark, and we got some meaningful results. It is a good basis for deeper research of this field in the future.