In Memoriam: Peter G. Neumann (1932-2026) – Communications of the ACM

Skip to content

Latest Issue

Search

Sign In

Join ACM

Peter Gabriel Neumann—known throughout the field as PGN—died on May 17, 2026, aged 93, from complications of a traumatic fall. For seven decades, he chronicled how computer systems fail and patiently advocated for the principles that make them fail less often.

Neumann was born on Sept. 21, 1932, in New York City. He entered Harvard Univerrsity in 1952 as a math major and took three Harvard degrees: an A.B. in mathematics (1954), and an S.M. (1955) and Ph.D. (1961) in applied mathematics. A Fulbright fellowship took him to Germany’s Technische Hochschule Darmstadt for a second doctorate, a Dr. rerum naturarum (1960). As an undergraduate, he took several graduate courses in computer-related topics, including Howard Aiken’s 1953 architecture course, and that summer, he programmed punch cards at the Naval Ordnance Laboratory. He dated his career as a computing professional to that year.

A single conversation shaped his intellectual life. In November 1952, while singing with the Harvard Glee Club at Princeton University, Neumann was invited to breakfast with Albert Einstein and spent more than two hours discussing complexity in mathematics, physics, cosmology, and, at length, music. Asked about Brahms, Einstein grimaced: “I have never understood Brahms. I believe he was burning the midnight oil, trying to be complicated.”1 Einstein’s dictum—that everything should be made as simple as possible, but no simpler—became, in Neumann’s own telling, the seed of every research program of his career.

Neumann worked at Bell Labs from 1960 to 1970, spending the first five years on error-correcting codes and survivable communications before becoming, from 1965 to 1969, a central figure in the Multics operating system project. With Bob Daley, he designed the Multics file system, including multilevel tree-structured directories, access-control lists, symbolic links, and backup mechanisms that modern systems still use. As one of the "triumvirate" running Multics with Fernando Corbató and Charlie Clingen, Neumann imposed the discipline for which Multics became known: No one was permitted to write a line of code until they had submitted a complete written English-language specification of the module’s behavior. Multics, Neumann later observed, solved the Y2K problem in 1965 and made stack buffer overflows impossible by construction.a

Neumann moved in 1971 to SRI in Menlo Park, Calif., where he worked for more than five decades as senior principal scientist and later chief scientist. Known for his warmth, kindness, and generosity, he fostered a worldwide community of researchers, historians, and journalists who work to catalog and publicize the risks of computing’s expanding role in daily life.

In 1976, when ACM SIGSOFT was formed, Neumann founded ACM SIGSOFT’s Software Engineering Notes (SEN), editing it for 19 years. A SEN column on computer-related mishaps spawned, in 1985, the ACM Risks Forum—known to its readers as comp.risks or the RISKS Digest, and at its peak, one of the most widely read mailing lists on the Internet. He moderated RISKS for 41 years, until April 2026, and curated more than 250 "Inside Risks" columns in Communications of the ACM.

Less than a week after he arrived at SRI, Larry Roberts of ARPA’s Information Processing Techniques Office (IPTO) visited, asking for a project on fault-tolerant computing; Neumann reframed the request as a trustworthiness problem, with fault tolerance as one aspect of it. The work fed into SRI’s Software Implemented Fault Tolerance (SIFT) project, which ran at NASA Langley for roughly 25 years and produced the foundational work on byzantine agreement by Lamport, Shostak, and Pease. From 1973 to 1983, Neumann led the design of PSOS, the Provably Secure Operating System: a tagged, capability-based, formally specified architecture. PSOS was never built; the era’s proof technology was inadequate.

In the mid-1980s, with his colleague Dorothy Denning, Neumann designed and built the Intrusion Detection Expert System (IDES). His involvement continued with Teresa Lunt’s NIDES and, from 1996, Phil Porras’s EMERALD. Those efforts are the conceptual ancestors of many current intrusion and anomaly detection systems. From 2010 on, as principal investigator of the SRI–University of Cambridge CTSRD project under DARPA’s CRASH program, he led work on CHERI (Capability Hardware Enhanced RISC Instructions), a clean-slate hardware-software co-design that revives capability-based protection at the instruction-set level. The architectural lineage was unmistakably from PSOS; CHERI has since moved from research prototype to commercial silicon, including Arm’s Morello platform.

Neumann was heavily involved in public policy beyond RISKS. He participated in four National Academies studies on computer security and cryptography: Multilevel Data Management Security (1982),...