[PRL] Fwd: Re: [Programming] Talk: "Fuzzing Abstract Interpreters" Dave Melski, Tues 1pm

[David Van Horn]

Here's the room info for Melski's talk(s) tomorrow.

David


-------- Original Message --------
Subject: 	Re: [Programming] Talk: "Fuzzing Abstract Interpreters" Dave
Melski, Tues 1pm
Date: 	Mon, 15 Nov 2010 16:02:53 +0000
From: 	Susan V. Welby <swelby at seas.harvard.edu>
To: 	Stephen Chong <chong at seas.harvard.edu>, EECS Programming List
<programming at eecs.harvard.edu>



*Dave Melski’s TALK*

The room for Dave Melski’s talk on “Fuzzing Abstract Interpreters” is on

Tuesday, November 16^th

*1:00pm in MD221*

*Dave Melski’s Guest Lecture in CS61*

Tuesday, November 16^th

*2:30pm in MD G-115*on reverse-engineering software

ALL Are Welcome to Attend!!

*From:*programming-bounces at eecs.harvard.edu
[mailto:programming-bounces at eecs.harvard.edu] *On Behalf Of *Stephen Chong
*Sent:* Sunday, November 14, 2010 9:29 AM
*To:* EECS Programming List
*Subject:* [Programming] Talk: "Fuzzing Abstract Interpreters" Dave
Melski, Tues 1pm

Hi all,
Dave Melski from GrammaTech will be giving a talk at 1pm on Tuesday Nov
16. Details are below. He is also giving a guest lecture in CS 61 at
2:30pm in MD G-115 on reverse-engineering software, which all are
welcome to attend.

Cheers,
Steve.

Title: Fuzzing Abstract Interpreters
Time: 1pm, Tuesday Nov 16,
Room: TBA
Abstract:

Sound static analysis tools have many applications, including assisting
with validation of other software. However, analysis tools must
themselves also be validated, and usually offer little assistance in the
way of “pulling themselves up by their own bootstraps.” Often the
problem is made more acute by the complexity of the algorithms needed to
achieve scalability and precision: attempts to implement such algorithms
may include implementation errors that lead to unsound results, no
matter how beautiful the theory backing the algorithms.

GrammaTech uses the /Transformer Specification Language/ /(TSL)/,
developed at the University of Wisconsin, to specify instruction-set
semantics and automatically generate machine-code analyzers. TSL greatly
simplifies the process of writing analyses. Perhaps as a consequence, we
have many analyzers that need validation. In this talk, I will describe
infrastructure we have developed to test the soundness of our semantic
specifications using fuzzing and comparison of abstract and concrete
executions. Obviously, testing cannot prove soundness, but this
technique has helped us to find many errors and increase our confidence
in our analyzers.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Attached Message Part
Url: http://lists.ccs.neu.edu/pipermail/prl/attachments/20101115/7d173fdc/attachment.txt