The Next Cybersecurity Battleground: Inside the Processor
Processors still execute instructions with no understanding of whether they are legitimate. Dover Microsystems’ CoreGuard, born from DARPA’s $100M CRASH program, adds a silicon-level enforcement layer that watches every instruction against programmable micro-policies. NXP engaged Dover in 2018 for embedded platforms. IBM finds 1 in 4 malicious breaches are now AI-enabled at $6M average cost. Dover holds 17 patents and has been validated through government red-team exercises.
Cybersecurity’s Location Problem
Cybersecurity faces a location challenge. Attackers often exploit vulnerabilities in systems businesses protect with software, yet the processor at the center remains largely unaware of any issues. Verizon’s 2026 Data Breach Investigations Report found vulnerability exploitation as the leading initial access vector, accounting for 31% of breaches, while only 26% of critical vulnerabilities were fully remediated during 2025.
Rising Stakes as Computing Deepens
As computing moves into physical systems, the stakes rise. AI workloads, connected infrastructure, industrial equipment, and autonomous machines all rely on processors to translate software instructions into real-world action. A compromised application could become more than a data-security problem; it can affect control.
IBM’s 2026 breach research found that one in four malicious breaches were AI-enabled, with those incidents costing an average of $6 million.
Software remains essential, but its complexity creates a persistent security challenge. Adding software defenses can leave organizations guarding vulnerable software with more vulnerable software. The question becomes: what if the processor itself could decide whether to execute an instruction?
Jothy Rosenberg, a 9x technology startup founder and Founder and Executive Chairman of Dover Microsystems, has spent years pursuing this question. His ventures include two exits exceeding $100 million, while his work at Dover centers on a proposition that challenges the software-first model of cybersecurity: the processor should have a mechanism to recognize prohibited behavior rather than blindly executing instructions.
“Processors are still the same simple basic design from 1945, which means they cannot tell if they’re being attacked,” explains Rosenberg. His concern is heightened by the fact that vulnerabilities remain even after security products are installed: “When you put this big system in place, and it’s supposed to protect you, and yet it has bugs, well, that’s just another avenue for the bad guys to get in.”
All software has been shown to have 15 bugs per thousand lines of source code.
Dover’s CoreGuard addresses the issue at the silicon level. The technology acts as an additional processor security layer that watches instructions as they are executed. Programmable rules, known as micro-policies, dictate what the processor is permitted to do. If an instruction violates a rule, CoreGuard prevents its execution and alerts the surrounding system.