Zhimin Gu

An online PPM prediction model for web prefetching

Web prefetching is a primary means to reduce user access latency. An important amount of work can be found by the use of PPM (Prediction by Partial Match) for modeling and predicting user request patterns in the open literature. However, in general, existing PPM models are constructed off-line. It is highly desirable to perform the online update of the PPM model incrementally because user request patterns may change over time. We present an online PPM model to capture the changing patterns and fit the memory. This model is implemented based on a noncompact suffix tree. Our model only keeps the most recent W requests using a sliding window. To further improve the prefetching performance, we make use of maximum entropy principle to model for the outgoing probability distributions of nodes. Our prediction model combines entropy, prediction accuracy rate and the longest match rule. A performance evaluation is presented using real web logs. Trace-driven simulation results show our PPM prediction model can provide significant improvements over previously proposed models.

A PPM Prediction Model Based on Stochastic Gradient Descent for Web Prefetching

PPM models are commonly used to predict the users next request in Web prefetching by extracting useful knowledge from historical user requests. Any of features such as page access frequency, prediction feedback, context length and conditional probability can influence on the prefetching performance of PPM models. However, existing PPM models only consider one or a few of them. Based on stochastic gradient descent, we present a novel PPM model that takes into account all the above mentioned features. Our model defines a target function to describe a nodes prediction capability, which is a linear combination of these features. In order to decrease the number of incorrect predictions, weights of these features are dynamically updated over every example according to the stochastic gradient descent rule. Our model selects a node with the maximum target function value among all matching nodes to predict the next most probable page. We use real web logs to examine proposed model. The simulation shows our model can significantly improve the prefetching performance.

Using Trust and Reputation Information to Choose Both Request Responders and Servers in Peer-to-Peer Networks

In current reputation-based trust models for P2P applications trust and reputation information are mainly used to choose service providers, they are helpless when facing this malicious attack: malicious nodes respond received queries at random in spite of whether it locally has queried resources. Once they are chosen as a provider, they will transmit a fake, even a malicious file. We present a hierarchical, especially reputation-based two-level trust model. In our model trust and reputation information are used to choose both request responders and service providers. So our model can restrain malicious behaviors from the headstream and withstand this attack. Furthermore in our model peers can receive more and better services; trust value in node can be concentrated at a considerable fast speed and malicious nodes can be identified quickly.

Selection of Training Period Based on Two-Window

Web prefetching is an important technique to reduce access latency. Although a number of prefetching methods have been proposed in the open literature, few of them study the length of the training period and most of them take arbitrary value for it. Therefore the prediction models of these approaches may keep a great deal of outdated information which occupies the high space and causes their prediction models low efficiency. Based on supervising the prediction accuracy, we present a two-window algorithm to decide the length of the training period. The large window indicates the training period. The small window tracks the prediction accuracy changing which decides whether older examples in the large window are considered outdated and forgotten. We use real web logs to examine proposed algorithm and the simulation shows that our algorithm can significantly improve the prefetching performance.

A reverse engineering approach of obfuscated array

Recently, research community has advanced in type reconstruction technology for reverse engineering. However, emerging with obfuscated technology, data type reconstruction grows more and more difficult, and obfuscated code is easier to be monitored and analyzed by attacker or hacker. Therefore, its essential to develop a novel approach to reverse engineering obfuscated array type based on refined CFG. We take split array for example and analyze feature in CFG and take advantage of compiler algorithm to identify obfuscated array.

Reconstruction of data type in obfuscated binary programs

Recently, research community has advanced in type reconstruction technology for reverse engineering, but emerging with obfuscate technology, data type reconstruction is difficult and obfuscated code is easier to be monitored and analyzed by attacker or hacker. Therefore, we present a novel approach for automatic establish data type inference rules and reconstruct type from obfuscated binary programs using machine learning algorithm.

Online granular prediction model for web prefetching

Web prefetching is a primary means to reduce user access latency. The PPM was used to predict user request patterns in traditional literature. However the existing PPM models are usually constructed in offline case, they could not be updated incrementally for user coming new request, such models are only suitable for the relatively stable user access patterns. In this paper, we present an online PPM granular prediction model to capture the changing patterns and the limitation of memory, its implementation is based on a noncompact suffix tree and a sliding window W, the results show that our granular prediction model gives the best result comparing with existing PPM prediction models.

Effectively Decreasing the Maintenance Overhead of Highly Dynamic Chord System

P2P systems are highly dynamic in nature. Nodes may join in or leave the P2P system at any moment. Frequently joining or leaving must increase the maintenance overhead greatly in DHT-based P2P system. The main reason of causing the cost is the lookup cost that nodes build their fingers. In this paper we introduce an iterative join algorithm for Chord that is suitable for highly dynamic environments. Iterative join algorithm builds the finger of node by iterative lookup and by the help of fingers information of nodes in the lookup path. Theory analysis and simulation show that Iterative join algorithm decreases efficiently the maintenance overhead and improve the lookup performance.

TsTlT: A Time-sensitive Two-level Trust Model Based on Reputation for Large-scale Peer-to-Peer Networks

Almost all of reputation systems for P2P applications are purely decentralized, trust management is complicated and convergence speed of trust in node is slow; even in most of them cost of network is overwhelming due to adopt broadcast searching mechanism. We present a partially decentralized time- sensitive reputation management model that has several advantages. Trust evaluator can define the valid period of trust evaluation based on own desire; Reputation system is efficient, convergence speed of trust is fast; Peers can receive more and better services; Furthermore it is scalable: network overhead is low; trust management is simple.