Lize Gu

A Semi-supervised Ensemble Approach for Mining Data Streams

There are many challenges in mining data streams, such as infinite length, evolving nature and lack of labeled instances. Accordingly, a semi-supervised ensemble approach for mining data streams is presented in this paper. Data streams are divided into data chunks to deal with the infinite length. An ensemble classification model E is trained with existing labeled data chunks and decision boundary is constructed using E for detecting novel classes. New labeled data chunks are used to update E while unlabeled ones are used to construct unsupervised models. Classes are predicted by a semi-supervised model Ex which is consist of E and unsupervised models in a maximization consensus manner, so better performance can be achieved by using the constraints from unsupervised models with limited labeled instances. Experiments with different datasets demonstrate that our method outperforms conventional methods in mining data streams.

Comment on Lee et al.'s group signature and e-auction scheme

Recently, Lee et al. used their new group signature with the function of authenticated encryption to design a sealed-bid auction scheme, and they claimed that their schemes are secure. In this paper, we show that if the group manager has a valid group signature of a member, without the member’s secret key, he can forge a group signature on arbitrary message on behalf of the member; then, if the registration manager (RM) and the auction manager (AM) conspired (with each other) in their auction scheme, they can forge a new bid on any goods on behalf of the bidder who has sent his/her bid to AM. Therefore, their group signature and auction scheme are insecure. Finally, we improve Lee et al.’s group signature scheme to overcome the modification attack and achieve the security requirements.

A correlation analysis method of network security events based on rough set theory

Network security event correlation can find real threat through correlating security events and logs generated by different security devices and can be aware of the network security situation accurately. This paper propose a network security events correlation scheme based on rough set, build database of network security events and knowledge base, gives rule generation method and rule matcher. This method solves the simplification and correlation of massive security events through combining data discretization, attribute reduction, value reduction and rule generation.

Improved tripartite concurrent signature

The concrete tripartite concurrent signature scheme designed by Susilo et al. is found being unambiguity. In other words, any fourth party can identify who is the real signer of the publicly available valid signatures before the keystone is released. In order to overcome this flaw, we proposed an improved concrete tripartite concurrent signature (iCTCS) scheme based on the ring signature. Every user can choose his keystone to guarantee the fairness in the proposed scheme. However, the performance analysis results show that the iCTCS scheme is existentially unforgeable under a chosen message attack in the random oracle, and satisfies correctness, ambiguity and fairness.

A new credential routing model based on trust theory

Router has acted as more and more important role in the large-scale network. In the meantime, the security of the router become more seriously. In this paper, we propose a new algorithm to decide a credential routing path from the start router to the end one based on the trust theory that considers the trustworthy of link, the cost of link and the trustworthy of router as the critical factors so as to utilize the principle of the grey correlation degree to obtain a credential routing path, instead of weighted mean in existed methods of trust measurement. We guarantee that packet can secure transfer under the credential routing path and identify the remote routers that are subject to attacks.