Earticle

Spam is a serious problem not only the number of floods but also more and more volatile type. It has caused a great impact on people's daily lives. Especially fraud spam, even cause huge losses to companies or individuals possibility. Therefore, it is imminent to filter spam efficiently. Existing spam filtering mechanism is mainly based on the character and content of the spam message. However, once the spam filter uses in other user's mailbox, the existing spam filtering techniques can not be well adapted. In this paper, we propose the adaptive spam filtering method for the above shortcomings. The method uses the unlabeled spam data that from other user or domain to enhance the adaptive and opposability of the anti-spam system. We use the transfer learning model to build the spam filtering system. A transfer learning model can use the untagged data, and migrate knowledge between different filter model, and improve the active collaboration of the filter.

Cloud computing appear to be a very popular and interesting computing technology. Every third person is using cloud computing directly or indirectly for example e-mail, most commonly used application of cloud computing, you can access your mail anywhere anytime. Your e-mail account is not visible on your personal computer but you have to access that with the help of internet. Like e-mail cloud computing provide many other services such as storage of any kind of data, access to different applications, resources etc. So users can easily access and store data with low cost and without worrying about how these services are provided to user. Due to this flexibility everyone is transferring data to cloud. To store data on cloud user has to send their data to the third party who will manage and store data. So it is very important for the company to secure that data. Data is said to be secured if confidentiality, availability, integrity is present. To secure data we have different algorithms. In this paper we will discuss the different cryptography of algorithms.

There is an inherent reliance on collaboration among the participants of mobile ad hoc networks in order to achieve the aimed functionalities. Collaboration is productive only if all participants operate in an honest manner. However, this is not always the case and these networks are subjected to a variety of malicious attacks. One of the key factors to ensure high communication quality is an efficient assessment scheme for node’s prediction trust, to choose potential cooperative nodes and reduce the probability of risk occurrence for next interaction. In this paper, firstly we propose a node’s trust assessment model based on node’s historical behaviors, in which the trust decision factors include the subjective reputation and indirect reputation. Then we try to combine an improved grey model with the Markov chain together to effectively predict the node’s trust. Experiment has been conducted to evaluate the effectiveness of the proposed mechanism.

As information technology environment gets based on mobile, cloud, BYOD, mobile terminal just used for communication in the past is changed to support e-commerce and mobile office using the internet. However, threat of tap and information leak due to terminal loss, theft or infection of malware occurs. Corporations are strengthening access control and security through attaching sticker to a smart phone camera but it looks not enough as fundamental protection measures. In order to maintain high level of security, security control function should be applied and location-based entrants’ security policy should be established so that a method to control mobile device can be directly suggested.

In the field of cloud security, the cloud provider don’t disclose any internal configuration information to protect itself, so the client know nothing about their data stored in the cloud and security status of the node providing services for them, thereby it causes the client’s worry whether to adopt cloud computing services. So that the trust between client and cloud computing provider become one of the biggest obstacles hindering the development of the cloud computing. Based on Direct Anonymous Attestation (DAA) and Dynamic Property Trusted Attestation (DPTA), we propose a client oriented remote attestation (CORA) model in cloud environment, client can select a node in the cloud at corresponding security level according to their own needs and can dynamically verify the node’s security status. At the same time, because the use of anonymous method it will not expose classified information of the node, such as configuration and location information etc. Furthermore we add service life of certificates to update certificates regularly, which enhanced the security of the attestation.

To protect digital content from illegal copy or reproduction, digital watermarking techniques are used which embed watermark into digital content and extract the same from the digital content to help in digital media protection. As techniques based on spatial and frequency domain are reported to have several limitations due to unsatisfactory values of fidelity by various researchers, new techniques based on genetic algorithm have been developed with an objective to optimize the values of fidelity of watermarked image. Genetic algorithms are used to find suitable locations for watermark insertion within a cover image, using either tournament selection approach or roulette wheel selection approach to provide optimization of fidelity. This paper is an attempt to provide a comparative study of the results obtained with genetic based watermarking techniques using roulette wheel selection approach and tournament selection approach . The variation of maximum fitness with respect to changing embedding strength, number of genes, mutation probability. Crossover probability and varying payloads has been compared and discussed for both the selection strategies.

In recent years, e-commerce has become one of the most active areas of scientific development, and more and more enterprises begin to build their own e-commerce platform. In this paper, we research the E-commerce capabilities by using factor analysis method; the result shows that electronic commerce can significantly improve enterprise performance. In this process, compound human resources, partner resources and information sharing will strengthen the role of e-commerce. In the era of e-commerce, enterprises can not only pay attention to their own business, the most important is to integrate their partners and resources, and only in this way enterprise be more competitive in market and to achieve win-win.

For issue that coal mining data tends to be incomplete, noisy and inconsistent, some popular classifiers are applied to predict class label of coal mining dataset. Noise and bad points are rejected from coal mining data which will be exchanged to input format suitable for mining. Then different classifiers are used to classify class label after extracting features. In the end, classification results are analyzed and knowledge assimilation is done. Experiment results show that decision tree model gives 88% of accuracy to correctly predict class label whereas neural network model predicts 85% correct class label. This research provides a powerful class label prediction tool as well as increasing knowledge of data classification models.