I have bundled everything up to deposit with OhioLINK's Electronic Thesis and Dissertation Center (ETD). Due to a cataloging issue involving Creative Commons licenses, OhioLINK has not yet posted the files.
Until this is worked out, you can find everything in the Harvard Dataverse.
Welcome, all! Here are the written materials for my talk today.
A Solder-Defined Computer Architecture for Backdoor and Malware Resistance
slides draft text
Marc W. Abel
Department of Computer Science & Engineering
Wright State University
2022 PDF
If you're visiting because you read my paper "Parallel Multipliers from Surface-Mount Static RAM," the code is here. It's written in Rust 2018. There are README and LICENSE files with the source code to answer your first questions.
If you're visiting because you read my paper in IEEE National Aerospace & Electronics Conference, the code is here, and an assembler manual is here. Neither has been stabilized for publication, but at least here is something if you need it today. You may either use this material under the terms of the GNU General Public License, version 3, or you may contact me to inquire about possible other licenses.
Today I’m going to tell you about a minicomputer you can build yourself with maker-scale assembly tools. I'll stress this word minicomputer. It's not a microcomputer, because it has no microprocessor, and no, there aren't any PLDs or FPGAs in the system either. Instead, the most complex logic chips in the base system are static RAM, and the this relative independence from semiconductor manufacturers will enable some of the most open, reliable, and secure computers in recent history.
Here are the slides for today's talk, “When Makers Make Secure CPUs.”
It is disheartening to dream of writing secure software—well organized, succinct, thoroughly validated operating systems and applications—knowing it would have to run on silicon with irreparable, undisclosed, and often deliberately introduced vulnerabilities. I propose a “supply chain firewall” for CPUs and systems that can shield (to a large extent) their purchaser or end user from the mistakes, misdeeds, and misaligned interests of semiconductor manufacturers.
A semiconductor plant costs 1 000 times as much as a pick-and-place assembly line, yet either can build a CPU. CPUs made in a “fab” are cheap in large runs, tiny, and perform computations at great speed. In contrast, CPUs soldered from smaller ICs allow superior process oversight during design and assembly, ability to inspect finished CPUs that does not exist with single-chip processors, affordability in even one-off lots, and more options with respect to assembly plant ownership and siting.
Here are the written materials for my talk today.
A Solder-Defined Computer Architecture for Backdoor and Malware Resistance
slides proposal
Marc W. Abel
Department of Computer Science & Engineering
Wright State University
2020 PDF
The reference ALU described in the above PDF has been modeled and tested in a C simulation. A tarball is available here.
This is a work in progress, but 109 opcodes do work now and have separately-designed regression tests. Some of the firmware does not exactly match the April document due to errors which have since been uncovered and corrected.
In the course of some graduate study, I have been researching the design of end-user-built CPUs, with particular emphasis on arithmetic logic units, or ALUs. The following resource is a snapshot of the work I have done to date, with a lot of technical intricacies to help newcomers to this technology come up to speed quickly. I will be giving a talk about this work in the near future, probably online on account of present conditions.
Elegant ALUs from Surface Amount SRAMs
Marc W. Abel
Department of Computer Science & Engineering
Wright State University
2020 PDF