Computer Science Colloquia
Friday, August 28th, 2015
Abbas Naderi Afooshteh
Advisor: Jack Davidson
Attending Faculty: John Knight (Chair), David Evans, Barry Horowitz (Systems Engineering) and R. Sekar (Stony Brook University)
2:00 PM, Rice Hall, Rm. 242
Ph.D. Proposal Presentation
Defeating Injection Attacks on Web Applications using Practical Threat Modeling and Hybrid Taint Inference
Research in taint-tracking techniques over the last decade has offered high hopes for thwarting injection attacks targeted towards Web applications. Despite numerous studies attesting to its effectiveness, taint tracking has not been widely adopted by the developer community. Impediments to adoption include relatively high performance overhead, inaccurate threat modeling, and deployment hurdles such as requiring administrator privileges. In contrast with taint tracking, taint inference infers taint markings based on either the program’s input (called negative taint inference), or the program itself (called positive taint inference). While taint inference avoids the negative aspects of taint tracking, current taint inference techniques suffer from multiple weaknesses and are not secure enough.
I propose new taint inference methods based on both dynamic and static analysis to enhance the security of current techniques so that they can be used to thwart injection attacks in general, and investigate them on layered threat and deployment models to effectively remove previous impediments to widespread deployment.