Six Degrees of RACF SPECIAL - Mainframe Attack Paths — Jonathan Prince

Join us at hack.lu 2026 — Info & Registration

Duration: 30 min

Type: Talk

Speakers: Jonathan Prince

Abstract

Enterprise identity security has been transformed by graph-based attack path modeling. BloodHound changed how attackers and defenders think about Active Directory, revealing privilege escalation paths that static access reviews consistently missed. That shift never reached the mainframe.
RACF, IBM’s security subsystem for z/OS, controls access to some of the most sensitive workloads in existence such as core banking, payment processing, government records, and insurance systems. Yet mainframe authorization is still primarily analyzed through flat reports and manual access list reviews, approaches that are blind to multi-step privilege escalation across users, groups, datasets, surrogate permissions, and system authorities.
RACFHound applies attack path modeling to RACF by transforming authorization data into an identity graph and integrating it into BloodHound via the OpenGraph framework. This talk demonstrates how privilege escalation paths through RACF can be systematically discovered using the same techniques now standard in Active Directory security. The mainframe becomes a first-class citizen in modern identity security analysis, and the attack paths hiding in plain sight in every large enterprise finally become visible.

Description

Modern red team and identity security tooling has largely ignored the mainframe. This is not because mainframes are secure, it is because the platform has historically been inaccessible to the wider research community, and security through obscurity has filled the gap. RACF configurations in large enterprises have often gone unaudited in any meaningful sense for years. The permissions are there. The escalation paths are there. Nobody has been looking.
This talk introduces RACFHound, a tool that models RACF authorization data as an attack graph and integrates with BloodHound using the OpenGraph framework. RACF entities are mapped to a graph model, and how attack primitives specific to z/OS emerge from that structure: SURROGAT permission chains, write access to APF-authorized libraries, escalation through UNIX System Services, and paths to the SPECIAL and OPERATIONS attributes. Live demonstration using BloodHound shows how paths that would require hours of manual RACF report analysis become instant graph queries.
The goal is to bring the analytical shift that BloodHound brought to Active Directory to an environment that handles more sensitive data, faces less scrutiny, and has fewer defenders who know what to look for.

View on pretalx