Maksym Nadtochyi

Maksym Nadtochyi

Auto-generated profile, based on publicly available information

Maksym Nadtochyi

Kharkiv National University of Radioelectronics,Department of Information Technology Security,Kharkiv,Ukraine

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 May 2026
Verified
Conference Paper
Closed access
Kushnerov OleksandrKushnerov Oleksandr
Kushnerov Oleksandr
VSVladyslav Sokol
Vladyslav Sokol
OAOlena Akhiiezer
Olena Akhiiezer
ACAndrii Cherniavskyi
Andrii Cherniavskyi
MNMaksym Nadtochyi
Maksym Nadtochyi
DBDmytro Balagura
Dmytro Balagura
The rapid transition from passive large language models to autonomous agent-based coding systems, including frameworks like Claude Code and OpenClaw, has significantly accelerated software development cycles while simultaneously complicating the enterprise attack surface. Conventional security paradigms relying on perimeter defenses and static role-based access control (RBAC) are inherently designed for deterministic entities, leaving them critically underequipped to mitigate threats posed by non-deterministic algorithmic agents. These autonomous systems frequently generate unpredictable, multi-step sequences of system calls upon ingesting poisoned external environment data or hidden prompt injections. This article proposes a novel Zero Trust architectural framework, complete with rigorous mathematical formalization, specifically engineered for secure multi-agent ecosystems. The architectural core features an intelligent Semantic Gateway deeply integrated with operating system-level coordination modules to dynamically provision ephemeral sandboxing environments. To achieve continuous authorization for asynchronous transactions, the framework utilizes a mathematical structure based on Heterogeneous Graph Neural Networks (HGNN). This algorithm continuously computes an agent's dynamic trust tensor by quantifying the topological divergence between its predicted and actual traversal paths within the deterministic space of a corporate repository graph. To empirically validate the analytical efficacy of this advanced architecture, the research team executed a large-scale simulation processing 100,000 asynchronous transactions. The proposed framework demonstrated exceptional predictive power, achieving a true positive rate (TPR) exceeding 98.7 % in the detection and preventive isolation of critical remote code execution vulnerabilities, including sophisticated CVE-2025-61260 exploits. Crucially, this robust security posture was maintained with a minimal false positive rate (FPR) of 1.1 % and an average processing overhead of just 45 milliseconds. Ultimately, this paradigm modernizes the NIST SP 800-207A standard by replacing static binary access controls with a continuous Markov decision process, transforming stochastic AI unpredictability into a quantifiable engineering asset for critical infrastructures.

11 December 2025 in Herald of Khmelnytskyi National University. Technical sciences
Verified
Journal Article
Open access
МНМаксим Надточий
Максим Надточий
ДБДмитро Балагура
Дмитро Балагура
ПШПавло Шулік
Павло Шулік
ВПВладислав Просолов
Владислав Просолов
Традиційні системи мережевої безпеки, що базуються на статичних правилах та сигнатурах, виявляються неефективними проти сучасних адаптивних кіберзагроз, зокрема Advanced Persistent Threats (APT), zero-day експлойтів та AI-генерованих атак. В роботі проведено систематичний аналіз відкритих наукових публікацій 2020-2025 років з метою дослідження проблеми недостатньої адаптивності існуючих систем безпеки та аналізу можливостей застосування штучного інтелекту (AI) та великих мовних моделей (LLM) для її вирішення. Дослідження підтверджують суттєве підвищення швидкості виявлення загроз при використанні AI-рішень у порівнянні з класичними підходами. Великі мовні моделі продемонстрували потенціал у галузі аналізу розвідки загроз та організації безпеки, однак постають питання їх стійкості до атак та інтерпретованості рішень.

30 September 2026 in Radiotekhnika
Verified
Journal Article
Closed access
MNM.M. Nadtochyi
M.M. Nadtochyi
DBD.S. Balagura
D.S. Balagura
Automating compliance analysis against catalogs such as NIST SP 800-53, which defines over a thousand hierarchical security and privacy controls, is attractive for resource-constrained organizations, yet sending sensitive infrastructure documents to cloud large language models is an unacceptable confidentiality risk. The natural alternative – local retrieval-augmented generation (RAG) on commodity hardware – raises a deployment question: given a privacy-mandated local RAG stack for NIST SP 800-53, which platform should an organization choose? Using a controlled, fully local protocol – three open-weight models (Llama 3.1 8B, Qwen 2.5 7B, Gemma 2 9B, all Q4) pinned to identical weights, 20 questions across four difficulty levels under three retrieval conditions, scored with RAGAS and a local judge for 180 measurements per platform per run, repeated three times – we report two findings. First, performance diverges sharply by whether a model fits in VRAM: the NVIDIA RTX 3060 is about 1.8× faster on the 7–8B models that fit, but for the 9B model 8 GB is no longer enough, part of the model spills to the CPU, and throughput collapses to 9.1 tok/s – nearly three times slower than on Apple Silicon (M1 Pro, 16 GB unified memory), which runs all three models at ~24–26 tok/s. Second, answer quality is formally equivalent across platforms under all conditions (TOST with pre-specified margin Δ = 0.05; p_TOST < 0.001 in every condition), so the platform decision reduces to a speed-versus-capacity trade-off rather than an accuracy one. The proposed hybrid retriever (regex identifier matching, metadata hard-filtering, and semantic search) with context injection yields a large, significant gain precisely on enhancement-level questions (level 4): ragas_avg +0.226 on Apple and +0.264 on NVIDIA, both p < 0.001, while leaving easier levels statistically unchanged.