Dark Matter

The hardware trust literature is overwhelmingly digital. Formal verification, equivalence checking, information flow analysis, Trojan detection benchmarks — all of it operates on logic. Meanwhile every real chip contains analog and mixed-signal blocks that the entire digital toolchain treats as opaque: phase-locked loops, data converters, power management, bias and reference circuits, I/O.

Dark Matter is the third entry in the lab’s Trojan series, after Dark Logic at RTL and Dark Fabric at the FPGA bitstream. It moves the question into the part of the design where the standard defenses do not reach — not because they fail, but because they were never pointed there.

The premise is uncomfortable and straightforward. A malicious modification to a clock generator, a reference voltage, or an ADC’s calibration does not have to change a single line of RTL to change what the digital logic computes. It does not appear in a netlist diff. It does not violate an equivalence check. Under nominal conditions it may not appear at all, surfacing only at a particular temperature, supply voltage, or process corner — which is also the region where legitimate analog behavior is hardest to distinguish from illegitimate.

Open questions

• What is the realistic payload of an analog or mixed-signal Trojan? Denial of service is easy; controlled, targeted corruption is the interesting claim.
• Which existing analog verification and test practices incidentally catch these, and which are structurally blind?
• Can detection be built from characterization data a foundry already produces, or does it require measurement nobody is currently taking?