Aerospace Cybersecurity Student Research Spotlight ERAU
Where
Aerospace Village
DEF CON 34
Duration
34 min
Presents two Embry-Riddle student prototypes: an AI log-triage system for aircraft cyber defense and a virtual satellite constellation testbed for security research.
Summary
This panel from Embry-Riddle's Center for Aerospace Resilient Systems covered two student projects built in the university's simulation labs. The first starts from a practical problem: aircraft logs are usually screened after flights with rule-based matching, which produces many false positives and a heavy analyst workload. The team built an offline language-model pipeline grounded in aviation documents, used it to create an aviation-specific attack framework rather than relying only on MITRE ATT&CK, and reported that it reduced manual review and handled about 2,000 logs per second on commodity hardware.
The presenters showed a cockpit prototype in Microsoft Flight Simulator. An ECAM alert flags affected systems, while a separate MCDU screen offers a summary and a generated procedure that the pilot can approve or reject. Repeated alerts are consolidated so they do not distract during critical phases of flight, and the team proposed a cyber degraded-law mode, similar to existing flight control fallbacks, for cases where systems must stay offline.
The second project, SpaceVE, models satellite constellations using virtual machines for flight hardware, NASA's core Flight System framework, GNU Radio links, and a ground-station scheduler. One demo traced an attack from a ground station through attitude control to an imaging satellite, where the corrupted output was only noticed after downlink. The takeaway is that researchers can study cross-domain effects, such as satellite faults reaching aircraft, without access to operational spacecraft.
Related talks
Main Stage·27 min
Ezz Tahoun
Shows how machine learning can group, deduplicate and enrich security logs before SIEM ingestion, cutting volume and stitching alerts into attack stories.
MEACS·32 min
Ezz Tahoun
Shows how on-prem machine learning can label, group, and chain SOC events into attack flows without dropping any log data.
MEACS·24 min
Ezz Tahoun
Shows how MITRE ATT&CK enrichment, clustering and deduplication can group SIEM alerts into attack chains while cutting log volume.


