EUVIMEDEUVIMED entdecken
Crossref

IEEE 754 STANDARD VULNERABILITIES IN CYBERSECURITY TASKS: ANALYSIS, CLASSIFICATION AND COUNTERMEASURES

T. H. Starushenko

Ukrainian Journal of Information Technology · 2026 · 8 (1) · 79-84

DOI: 10.23939/ujit2026.01.079 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Ukrainian Journal of Information Technology, 2026. Verfasst von T. H. Starushenko. Die Arbeit nennt 10 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The cybersecurity implications of numerical anomalies inherent to the IEEE 754 floating-point arithmetic standard remain insufficiently studied, despite their confirmed role in critical software vulnerabilities across embedded systems, cryptographic libraries, and web applications. The special-value semantics of IEEE 754-2019 – including Not-a-Number (NaN), signed zero, positive and negative infinity, and denormalized (subnormal) numbers – were investigated as potential attack vectors and vulnerability amplifiers in real-world computing systems. A total of 312 entries from the NIST National Vulnerability Database (NVD) covering the period 2015–2024 were analysed, categorized by the underlying IEEE 754 anomaly mechanism. A controlled numerical experiment was conducted to quantify rounding-error accumulation across 1000 iterations of summation in binary32 and binary64 formats, as well as with Kahan compensated summation, which demonstrated a difference of four to five orders of magnitude between formats. Six categories of IEEE 754-related security threats were identified and classified: NaN-comparison bypass attacks, timing side-channel leakage through denormal-number processing on x86 and ARM platforms, floating-point overflow and underflow exploitation, systematic rounding-error accumulation in iterative algorithms, signed-zero inconsistency attacks targeting hash tables and comparison logic, and unsafe numeric type-conversion vulnerabilities. The practical relevance of each category was established through documented real-world incidents: CVE-2023-36664 (Ghostscript, CVSS 9.8), CVE-2022-26134 (Atlassian Confluence, CVSS 9.8), CVE-2021-3156 (sudo, CVSS 7.8), an OpenSSL timing side-channel, and the Ariane 5 Flight 501 catastrophic failure caused by a 64-bit float-to-16-bit integer conversion overflow. A unified taxonomy was developed that links each threat category to real CVE records and to five countermeasure strategies – explicit NaN/Inf validation, safe floating-point arithmetic libraries, constant-time implementations, formal verification tools (Frama-C, Fluctuat, Why3), and static analysis (SAST) instruments – with a multi-criteria comparative evaluation across effectiveness, implementation cost, compatibility, and deployment speed. The proposed classification provides a structured foundation for secure software audits, penetration testing of floating-point subsystems, and the formulation of IEEE 754-aware secure coding guidelines compliant with MISRA C and CERT C standards.

Zitieren

APA

Starushenko, T. H. (2026). IEEE 754 STANDARD VULNERABILITIES IN CYBERSECURITY TASKS: ANALYSIS, CLASSIFICATION AND COUNTERMEASURES. Ukrainian Journal of Information Technology, 8(1), 79-84. https://doi.org/10.23939/ujit2026.01.079

Vancouver

Starushenko TH. IEEE 754 STANDARD VULNERABILITIES IN CYBERSECURITY TASKS: ANALYSIS, CLASSIFICATION AND COUNTERMEASURES. Ukrainian Journal of Information Technology. 2026;8(1):79-84. doi:10.23939/ujit2026.01.079

BibTeX
@article{Starushenko2026,
  author = {Starushenko, T. H.},
  title = {IEEE 754 STANDARD VULNERABILITIES IN CYBERSECURITY TASKS: ANALYSIS, CLASSIFICATION AND COUNTERMEASURES},
  journal = {Ukrainian Journal of Information Technology},
  year = {2026},
  volume = {8},
  number = {1},
  pages = {79--84},
  doi = {10.23939/ujit2026.01.079}
}
Kostenlos starten →