Sweeney_cmu_0041E_10634.pdf (16.58 MB)
Download file

Obfuscation and Security for Digital Integrated Circuits

Download (16.58 MB)
thesis
posted on 10.06.2021, 21:01 by Joseph SweeneyJoseph Sweeney
Globalization of IC manufacturing has led to increased security concerns, notably IP theft. A promising countermeasure is logic locking that adds programmable elements to a design, obfuscating the true functionality during manufacturing. Generally, logic locking techniques aim to provide IP security while avoiding large overheads. Towards this end, this dissertation makes several contributions. First, a security analysis of existing locking techniques is presented, exposing several vulnerabilities. One class of techniques is analyzed using sensitivity, a property of Boolean functions. The analysis reveals the modified portions of a circuit with high probability, leading to deobfuscation. Another class of locking methods is used to
demonstrate two modeling techniques, relaxed models and symmetry breaking, that can dramatically reduce attack times. These vulnerabilities inform the development of latch-based logic locking, a novel obfuscation method that resists known attacks while maintaining low overheads. This
balance is achieved by locking a design’s clock tree, manipulating the functionality while avoiding timing-critical logic. To validate the technique, a set of common industrial designs has been locked and brought through the full manufacturing process. To demonstrate resistance to deobfuscation, the locking scheme is evaluated against existing and newly developed attack methods. Finally, two metrics are established to better quantify the security of a given locking technique under common attack scenarios. These metrics are efficiently estimated using approximate model counting techniques. Importantly, they provide a means of analyzing the overhead-security trade-off of locking techniques, an essential aspect of integrating locking schemes into real systems.

History

Date

25/02/2021

Degree Type

Dissertation

Department

Electrical and Computer Engineering

Degree Name

  • Doctor of Philosophy (PhD)

Advisor(s)

Lawrence Pileggi

Usage metrics

Exports