Crash2Compromise: CrashGuard-FS for Security-Aware Linux File Updates

Many Linux programs assume a successful write() or close() constitutes a committed security update, but directory and file contents may remain buffered under the application boundary. We reconceptualize crash-consistency as a cyber-attack when the targeted object is an SSH policy, sudo rule, EDR configuration, package database, credential-metadata file, or audit log – making a crash into an escape vector to defense-loss, evidence-obfuscation, authorization-drift, and recovery-panic. Our CrashGuard-FS framework uses a temporary-file, fsync(), rename(), and parent-directory fsync() sequence to secure operations, complemented by security classification, hash-chained evidence, and crash-point analysis. A simulated crash-point framework evaluates five update sequences for their effectiveness at securing six security-sensitive artifacts. Direct overwrite introduces an 86.5% per-operation security failure rate on average, compared to 11.2% per operation with a total file-and-directory persistence protocol and 5.2% per-operation in our modeled CrashGuard-FS. We did not build a file system; we built a security engineering technique that makes crash-sensitive Linux updates inspectable, auditable, and defensible.