Choose at most three.
That cap is deliberate. Spring 2024 ran binary reverse engineering, buffer overflow, and the timing side channel. A fourth module in the same term starts to feel like a second course.
For instructors
Start from the design topic on your syllabus, then pick the module that makes a student use it. Do not start from a list of famous attacks and look for a week to put them in.
That cap is deliberate. Spring 2024 ran binary reverse engineering, buffer overflow, and the timing side channel. A fourth module in the same term starts to feel like a second course.
Plan 30 to 40 minutes of new lecture for each module. The student work replaces an existing homework or lab, it does not stack on top of one. For embedded systems, the natural slots are the weeks on the toolchain, the calling convention, and the timer and serial port.
M1 through M3 need Keil and an AT89C51 target or a faithful simulation of it, plus the logic analyzer built into the debugger for M3. M4 needs a populated board and a multimeter, and a logic analyzer only for the optional protocol check. M5 needs a browser. M6 through M8 need a VHDL simulator. M9 needs a student laptop.
On M2 and M3 the second half is the point: a bounded copy, and a constant-time comparison measured again. On M6 and M8 the lock itself is the work. If you make the module extra credit, expect some students to skip it and then meet it on the exam.
Taking the board apart needs the lab. Reading datasheets and drawing the block diagram does not. The same split helps M3 when setting up the logic analyzer takes the whole period.
Embedded systems
This is the sequence that has been taught and surveyed. Run them in that order. M2 assumes the student can already read the memory windows from M1. M3 assumes the timer and the serial port.
Digital logic
Students build the adder and lock it, then receive that locked design and add a Trojan without touching the lock. M8 can replace M7 if the course is about to teach state machines. M5 can stand alone in the week on CMOS power.
Computer architecture
Two parts. The first reads the machine’s own cache geometry. The second treats a hit and a miss as a channel. There is a recorded measurement and no separate graded write-up in the archive.
Student handouts, slide decks, starter VHDL, the C sources for the embedded labs, and the binary used in the reverse-engineering homework. Lab manuals for the three VHDL experiments are included as documents because they carry the figures.
Instructor guides, reference VHDL with completed Trojans and completed locks, and the solved homework from earlier offerings stay in the project archive. Lecture recordings from Spring 2023 are in that archive too. Ask the KU group for the instructor packet.
Two further exclusions. The published tutorial on obfuscation and logic locking is cited from the VHDL labs and is not redistributed here. A gem5 cache configuration script that ships beside the cache slides is third-party teaching code and is not copied onto this site.
This work was supported in part by the U.S. National Science Foundation (NSF) Secure and Trustworthy Cyberspace (SaTC) program under Award No. 2114157.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the U.S. National Science Foundation.
The project is a collaboration of the University of Kansas, the University of Florida, and Florida International University.