Research On Software Protection Technology Based On Watermark And Feature

With the growing popularity of network services, software infringement, piracy, tampering and malicious attacks are also increasing, seriously impeding the healthy and sustainable development of the software industry. Software protection technology has received much attention from governments, entrepreneurs and scholars. Consequently, software protection techniques like copyright management, piracy prevention, software recognition, etc., have become hot and key research problems.Currently, a number of software protection techniques exist, including hardware-based protection like softdog, FLEXlm and SmartCard, software-based protection like watermark, birthmark and encryption techniques, and protection methods based on the binding of hardware and software. Software watermarking belongs to digital watermarking. In recent years, researches on software watermarking protection increase continuously,because watermarking is more concealed than other techniques like softdog, has one more layer of protection than encryption, and can provide legal proof. However, these researches mainly focus on the copyright protection of text and multimedia, while researches on software watermarking and database watermarking are relatively less. So the concerned studies are still faced with many challenges. Software feature recognition (also known as birthmark or zero watermark) is to extract the invariant features (or key features) from software, for identifying the software or the software family. If two pieces of software have a greater degree of similarity in their birthmark features, one of them is likely to be a piracy of the other’s, or they two belong to the same family. Feature-based detection and identification gains great popularity because it doesn’t degrade the performance of software. However, researches on software feature recognition are still in their initial stage.Many software watermarking and birthmark algorithms have been proposed, yet they can not meet the application requirements concerning their security, robustness, credibility, etc. Protection system based on watermarks and features shall adapt to the development of the network environment, as well as the new demands of multiple users and multiple permissions. Software watermarking is still facing such attacks as cutting, twisting, adding and conspiracy. The accuracy of feature detection is low, so it can not represent or detect software accurately. Therefore, the study of software protection and recognition based on both watermarks and features are launched in this dissertation, to provide theoretical analyses and technical supports for building a safe software protection system.The main contents of the thesis are summarized as follows:1. A copyright protection protocol is proposed based on multi-watermarking.Researches on multiple watermarking mainly appear in the fields of image and video protection. This thesis analyzes the deficiencies of single watermarking in software protection and discusses the coupling method of multiple watermarking, and then, put forward a protection protocol for version role constraints based on multiple watermarking. This protocol separates effectively the copyright verification and authentication through multiple watermarking, to meet the needs of multiple versions and multiple users, and is suitable for network application environment. Studies are made on the processes based on multiple watermarking, such as watermark loading, copyright registration, online transaction, digital production validation. These studies provide explorations on the copyright protection of software for network demands and security services, and have certain practical value.2. A software watermarking algorithm based on threshold scheme is put forward.Considering the two types of attacks at software watermarks, secret sharing ideas in cryptography—(t,n) threshold scheme, is applied to watermarking. Information carried by the watermarks can be recovered only by extracting t (t<<n) sub-watermarks. Based on dynamic graph software watermarking, this thesis studies the system of Asmuth-Bloom threshold scheme, goes into the embedding and extracting processes of threshold scheme watermarks, and improves the dynamic graph watermarking. A large watermark is divided into several sub-watermarks and stored into different positions in the program. Finally, analyses and experimental comparisons are conducted on the watermarking algorithm which is demonstrated to have certain advantages against some specific attacks.3. A software watermarking algorithm based on chaos optimization is proposed.Software watermarking based on threshold scheme, conditioned that the sub-watermarks are not completely destroyed, has some limitations. In view of this deficiency, chaos theory is applied in the study to improve it by enhancing the robustness and concealment of each sub-watermark. The study deals with chaos theory and chaotic system and their applications in digital watermarking, and improves watermarking through chaotic hash and chaotic encryption. The study focuses on watermarking on the basis of executable PE files, and considers confusing the embedding position of sub-watermarks and increasing their resistance to reverse attacks through chaotic encryption. Finally, prototype system of chaos optimization is accomplished. Experimental analyses and comparisons demonstrate that the robustness and concealment of software watermarking can be improved by this system.4. An algorithm of birthmark feature selection based on cluster analysis is proposed.In light of the existing problems of credibility and redundancy of birthmark features, a birthmark feature selection method based on cluster analysis is put forth to improve conventional segmentation or slice selection methods. This selecting method considers the change of features and the relations between them when software changes. In this method, information are used to measure the distance between classes and to construct the gain functions and penalty functions, and the correlation coefficient is used to measure the distance within a class and to reduce the redundancy of within-class features. Extraction and clustering are carried out based on dynamic sub-behavior features. Generalization to other types of features based on their similarities is also studied, to make this algorithm fit for software birthmark feature selection of different types (string, vector, collection, tree and pattern), providing an effective way for feature filtering.5. A software recognition algorithm based on boosting multi-attribute feature is set forth.For the low accuracy of single-attribute feature detection based on threshold value in software recognition, a software recognition algorithm based on weighted boosting multi-attribute feature is proposed. Classification learning and ensemble learning are addressed, as well as the measurement and description of multi-attribute features. Two static features and two static "behavioral" features are extracted, measured and learned, and finally weighted-voted according to accuracy rate, in this way the result of unknown software recognition is given. Sequence segmentation and characteristic graph measurement in feature selection are resolved, and robustness and credibility of features are improved, consequently the accuracy of software recognition is enhanced.