Classify and Rank Daikon Invariants on the Minicar Pervasive Computing Platform

MPhil Thesis Defence


Title: "Classify and Rank Daikon Invariants on the Minicar Pervasive Computing 
Platform"

By

Mr. Chunlin Zhu


Abstract

Dynamic invariant inference derives likely program properties based on observed 
variable values from concrete program executions. It has emerged as a highly 
promising software engineering practice recently. Among various dynamic 
invariant inference tools, Daikon is the first and the most mature 
representative with the widest use in various applications. However, Daikon’s 
inferred invariants suffer from irrelevant ones seriously, as pointed out by 
DySy, about 80% of Daikon’s invariants are considered irrelevant by human 
users. In addition, Daikon’s existing C-language front ends are mainly used on 
Intel-386 compatible computers. They do not support program execution trace 
recording on the Minicar platform, which features some typical pervasive 
computing characteristics and is used as our test bed. To address these 
difficulties, we build a new Minicar-specific C-language front end for Daikon. 
We then propose to classify the invariant results, and rank a method’s inferred 
invariants based on their containing variables’ relevance calculated by the CRF 
models according to the function execution in this thesis. In this way, we 
distinguish relevant invariants from irrelevant ones and place a method’s most 
relevant invariants at the front of the result. The experimental results show a 
significantly improvement on the presented invariants’ degree of relevance.


Date:			Friday, 17 July 2009

Time:			10:00am - 12:00noon

Venue:			Room 3501
 			Lifts 25-26

Committee Members:	Dr. Shing-Chi Cheung (Supervisor)
 			Dr. Charles Zhang (Chairperson)
 			Dr. Sunghun Kim


**** ALL are Welcome ****