@techreport{oai:ipsj.ixsq.nii.ac.jp:02006941,
 author = {Shengyang,Li and Kenji,Kono and Shengyang Li and Kenji Kono},
 issue = {18},
 month = {Feb},
 note = {Rust ensures strong memory safety through ownership and borrowing; yet, these guarantees often collapse at the cross language boundary. This transition creates a semantic gap―comprising vertical semantic loss during compilation and horizontal context discontinuity at the language boundary―leaving the Rust compiler oblivious to native execution. Existing analysis tools struggle to bridge this gap, suffering from either semantic erasure during lowering or prohibitive overhead from exhaustive instruction-level monitoring. We present RINA (Rust Invariant and Native Auditor ), a semantic-aware auditing framework that utilizes WebAssembly (Wasm) as a unified intermediate platform. RINA's key insight is to treat cross-language vulnerabilities as semantic invariant violations detectable at language boundaries. It statically lifts rich Rust semantics (e.g., ownership markers and thread-safety traits) as metadata and dynamically enforces these safety contracts via boundary-centric binary instrumentation. Preliminary evaluation on 21 real-world FFI CVEs across memory, type, semantic, thread, and exception safety dimensions demonstrates that RINA achieves a 95.2% recall rate (20/21) with low runtime overhead. RINA provides an efficient and practical solution for maintaining Rust's safety guarantees in an increasingly inter-connected ecosystem., Rust ensures strong memory safety through ownership and borrowing; yet, these guarantees often collapse at the cross language boundary. This transition creates a semantic gap―comprising vertical semantic loss during compilation and horizontal context discontinuity at the language boundary―leaving the Rust compiler oblivious to native execution. Existing analysis tools struggle to bridge this gap, suffering from either semantic erasure during lowering or prohibitive overhead from exhaustive instruction-level monitoring. We present RINA (Rust Invariant and Native Auditor ), a semantic-aware auditing framework that utilizes WebAssembly (Wasm) as a unified intermediate platform. RINA's key insight is to treat cross-language vulnerabilities as semantic invariant violations detectable at language boundaries. It statically lifts rich Rust semantics (e.g., ownership markers and thread-safety traits) as metadata and dynamically enforces these safety contracts via boundary-centric binary instrumentation. Preliminary evaluation on 21 real-world FFI CVEs across memory, type, semantic, thread, and exception safety dimensions demonstrates that RINA achieves a 95.2% recall rate (20/21) with low runtime overhead. RINA provides an efficient and practical solution for maintaining Rust's safety guarantees in an increasingly inter-connected ecosystem.},
 title = {RINA: Semantic-Aware Enforcement of Rust Safety Invariants at Cross-Language Boundaries},
 year = {2026}
}