The paper presents a formalized model of the architecture of a software system that provides confidential access to the information resources of computer networks amid the high dynamism of network infrastructures, the increasing complexity of the information environment, and the heightened requirements for protecting user privacy. The proposed architecture is grounded in the concepts of modularity, openness, and unified software interfaces, which together ensure compatibility with heterogeneous deployment environments. The functional subsystems integrate multi-level authentication, cryptographic encryption, attribute-based access control, anomaly detection in user behavior, and active access monitoring based on artificial-intelligence mechanisms. A mathematical model for assessing the degree of confidentiality of access has been developed, enabling quantification of the security level of information objects in accordance with Zero Trust Architecture standards. The practical significance lies in the architecture's applicability to information systems with stringent confidentiality demands while meeting institutional and regulatory oversight requirements.
The growing number of cybersecurity incidents accompanying the digital transformation of enterprises requires new approaches to organizing secure access to the information resources of computer networks. Traditional solutions based on centralized access control or classical cryptographic schemes are insufficiently effective in distributed, multi-service environments, where the volume, speed, and variability of traffic demand dynamic, flexible, and adaptive protection.
Of particular importance are software-based confidential-access systems that can provide not only authenticated and authorized access but also preserve the confidentiality of user requests, their sessions, and the history of interactions with network resources. In modern implementations, such systems should support a multi-level security policy, automated access control, distributed architecture, standardized encryption protocols, and interaction with other components of the information infrastructure—including SIEM, VPN, and Zero-Trust environments. These systems must offer open, unified interfaces; a flexible, modular architecture; and the capability to integrate with SIEM, DLP, and VPN solutions, audit tools, digital-identity systems, and secure-transmission protocols (TLS, IPsec, WireGuard, etc.).
The purpose of this paper is to develop an architecture for a software system that enables confidential access to the information resources of computer networks, integrates security mechanisms with artificial-intelligence tools, supports information-security standards, and allows access policies to adapt to the evolving risk landscape of IT infrastructures. The study proposes a system model, a mathematical apparatus for assessing the level of confidentiality, and principles for building an open architecture that accommodates third-party software components.
The issue of designing the architecture of software systems that enable confidential access to
information resources is a key topic in contemporary cybersecurity research, where special
attention is devoted to adaptive risk-management mechanisms, data protection, and dynamic
access control. Althar et al. [
Wang, Ahmad, and Bakar [
The study by Barraza de la Paz et al. [
The review by Ahmad et al. [
Thus, contemporary scientific literature confirms the relevance of creating architectures that
support flexible interaction among security components, adaptive access control, and the
confidentiality of information processing—factors that collectively form the foundation for
developing effective software systems for confidential access in computer networks. Recent
research by Vakhula et al. [
The research methods are grounded in the theory of algorithmic computational complexity, the reliability of automated control systems, modern cryptographic primitives, distributed programming, and the theory of multi-layer architectures. Additionally, risk-based modelling in accordance with ISO/IEC 27005 and NIST SP 800-30, together with machine-learning techniques for adaptive query analysis, were employed. Formal models of confidential access, fuzzy-logic methods, Bayesian updating, and agent-based interaction within a distributed environment are applied. The combined use of state-of-the-art cryptographic protocols (TLS 1.3, IPsec, SSH-2), ZeroTrust concepts, and integration with SIEM and MFA systems provide an adaptive, scalable architecture for confidential access to critical information resources.
Modern software tools for protecting electronic information can be classified into the following
groups: cryptographic protection tools, access-control systems, tools enabling confidential access to
information resources, and security mechanisms deployed in financial, educational, and cloud
platforms [
In this context, particular attention is devoted to data-protection solutions based on the IPsec protocol, which provides confidential and authenticated exchanges in TCP/IP networks. Modern mechanisms for building Virtual Private Networks (VPNs) are also examined; they allow secure data exchange between distributed information systems while minimizing infrastructure costs and simultaneously broadening the geographical reach of access [7, 8]. Notable effective implementations include IPsec VPN, SSL VPN, and tunnelling with WireGuard.
Figure 1 illustrates the deployment of the components of a software-based confidential-access
system designed in accordance with Zero Trust security principles and contemporary authorization
and encryption standards. The architecture comprises a client device, authorization and access
servers, a cryptographic gateway, a security-information and event-management (SIEM) system,
and cloud services hosting educational, financial, and other resources. The inter-component
connections demonstrate the flow of access requests, authorization decisions, cryptographic
protections, event logging, and configuration policies. The system supports scalability, adaptability,
and integration with external platforms through unified interfaces.
which no trust is granted to any user or device by default [
Typical architectures of such systems are characterized by modularity, where the key
components are: a module for collecting and analyzing event information, a decision-making
module based on security policies, a response module that implements blocking or allowing access,
a logging module that records the actions of access subjects for auditing, and a management
module that supports dynamic policy configuration in line with environmental changes [
The component architecture of the software system for confidential access to the information resources of computer networks (Figure 2) is built with due regard for modern data-protection requirements. It includes modules for multifactor authentication, policy-based authorization, and access control in accordance with the Zero-Trust concept. The system features cryptographic protection (IPsec, TLS, SSH-2, WireGuard), event logging, incident response, and monitoring components through integration with SIEM platforms. Access to cloud resources—including educational and financial services—is provided through secure confidential tunnels. The architecture supports dynamic access-rights management, end-to-end encryption, user identification, and traffic protection in accordance with industry information-security standards.
Special attention is devoted to the construction of confidential communication channels. Their
organization includes key exchange via the IKEv2 protocol, the use of Encapsulating Security
Payload (ESP) for traffic encryption, and Authentication Header (AH) for source verification [
The study examines algorithms and architectures for confidential access to information
resources, in particular technologies of anonymous virtual networks such as DC-net, Mix-net,
Crowds, Onion Routing, and Freedom Network [
The generalized architecture of such systems includes modules for authentication,
authorization, response, logging, access-policy management, and traffic encryption using modern
cryptographic protocols (TLS 1.3, SSH-2, IPsec with PFS) [
The architecture of the anonymous Mix-net network was reviewed to deepen understanding of
the principles underlying confidential-access channels in distributed information environments
[
The practical value of this analysis lies in adapting the anonymization principles implemented
in Mix-net to the architecture of the system under development, which delivers confidential access
to information resources [
The diagram (Figure 3) depicts the complete sequence for routing an anonymous user request to a secure resource over the Mix-net network, incorporating Java Anonymizer Proxy (JAP) as the initial proxy service. In the first stage—Anonymous Request Routing—the user submits a request through JAP, which applies the first layer of encryption. The request then proceeds to Mix Server 1, where it is re-encrypted (Layer 2), is forwarded to Mix Server 2 for another encryption layer (Layer 3), and finally reaches Mix Server 3, which performs final processing and relays the request to a secure resource, such as a financial, cloud, or educational system. The second stage—Response Routing—traces the return path from the resource back to the user: the response travels through Mix Server 3, where it is again encrypted, passes through Mix Server 2 and Mix Server 1 with corresponding re-encryption layers, and is ultimately received and decrypted by JAP for the user. The diagram illustrates the full cycle of request and response processing within a secure environment that adheres to principles of end-to-end encryption, route obfuscation, multi-layer encryption, and source-anonymity protection. The combination of these approaches in a single scenario demonstrates the practical implementation of a confidential-access architecture that not only safeguards traffic at every stage but also effectively counters correlation attacks and traffic analysis, thereby maintaining the privacy and integrity of information exchange in computer networks.
The study proposes a formalized mathematical model and appropriate methods for
implementing confidential access in computer networks, focused on the protection of critical
information resources [
The information system is modeled as a set of objects: (2) (3) (1) (4) where each oi is an information resource that stores or processes data. Set O represents all objects of the information system, which can be either stored data or components that process information (e.g., databases, services, files).
Entities interacting with the system: where each s j—user, device, or software agent. Set S describes the subjects of the information system, i.e. everyone who can access objects: users, devices, processes, agents.
Each object and subject is assigned a security rank from a partially ordered set:
O = {o 1 , o 2 , … , o n}, S = {s1 , s2 , … , sn}, g : S × O →R ,
Set R —are security ranks that are ordered by the degree of trust. They are used to categorize both subjects and objects by level of sensitivity or trust. For example: r 0—public information, r k— the highest level of confidentiality.
As for the access control policy model, the access ranking function is used: Function g ( s , o ) defines the level of security required for an entitys to gain access to an object o . This mapping is used to determine the minimum subject rank required to access the corresponding object in the system. In other words, it reflects the minimum rank that a subject must have in order to access an information resource.
Maximum access level of the subject:
Determines the highest level of access that subject s has among all the objects that he or she can access. This allows the system to determine the general level of privileges of this user.
Minimum level for an external attacker: rang ( s )= {g ( s , o )∨ o ∈ O },
rang ( s attacker )= { R } =r 0,
For an external attacker, the lowest possible access level is set to —r 0, which means no trust. This is the starting point for analyzing potential threats.
The access decision is formalized as a binary function:
Access ( s , o )= { 1 , if rang ( s ) ≥ rang ( o ) 0 , otherwise , (7) where rang ( s )= {g ( s , o )∨ o ∈ O }, а g : S × O →R is a ranking function. It is a binary decisionmaking function for granting access. If the security level of the subject is not lower than the security level of the object, access is allowed (1), otherwise—is denied (0).
The model for assessing the level of confidentiality over time is based on the exponential decrease in the probability of maintaining data confidentiality: (5) (6) (8)
C (t )= P secure (t )= e− λt, where λ is the intensity of the risk or attack, and t is the time of data retention. The model reflects that over time, the probability that information will remain confidential without additional protection measures decreases exponentially. The presented function is an important tool for formalized analysis of the loss of confidentiality in dynamic information systems. Thus, the proposed model allows us to mathematically describe privacy dynamics, adapt access policies to the time characteristics of data storage, and increase the efficiency of confidential access architectures in computer networks.
To build a more flexible protection system, the concept of a multilevel security model is
introduced, in which each component of the IT system is assigned a specific protection level [
The diagram (Figure 4) illustrates the logic of the software system that provides confidential access to the information resources of computer networks. The system performs access control by comparing the security ranks of subjects and objects, which are determined during the ranking process. Access requests are processed in accordance with established policies that incorporate security parameters and current threat models. Decisions to grant or deny access are logged and analyzed by the SIEM system to detect anomalies or unauthorized actions originating from either legitimate users or potential attackers. The diagram depicts the interaction among the user, access controllers, policies, data objects, and monitoring systems.
The concept of a multi-level information-security model for automated data-processing systems
involves structuring system components according to their protection level and functional purpose.
This model enables clear delineation of areas of responsibility, minimizes the risk of compromise,
and supports the application of adaptive security policies at different logical tiers. In particular,
Level A encompasses users’ internal client devices; Level B protects an isolated information
resource (for example, a web application that processes confidential data); Level C comprises the
network infrastructure that supports internal interaction; and Level D represents external users
and the channels through which they access the system via public networks—most notably, the
Internet [
Within the confidential-access model, all system principals are divided into the following logical
groups: session initiators—users or devices that request resources; relay servers—nodes that
obfuscate traffic routes (e.g., Mix servers or Tor nodes) [
In the proposed architecture of the confidential-access software system, data are transferred in
accordance with cryptographic-protection principles, ensuring confidentiality, integrity, and
authenticity at every stage of the route. Repeaters (intermediate nodes that relay encrypted packets
between the client and the target server) employ symmetric-encryption keys, enabling rapid
processing of large traffic volumes with minimal latency [24]. Symmetric encryption is effective for
tunnelling information when keys are pre-shared or transmitted over a secure channel. Receiving
servers (route endpoints) use asymmetric cryptography—specifically, public-key mechanisms—
which allows them to accept session keys or user requests securely without a pre-established
shared secret [
K session= F ( x , k ),
(9)
where x is the random variable (nonce or initialization vector) that ensures the uniqueness of
each connection, k is the encryption key, which can be symmetric or obtained by exchanging
through asymmetric methods (for example, through ECDH), F ( x , k ) is the cryptographic key
calculation function (e.g., HMAC, PBKDF2, or HKDF) [
To build an anonymous channel, a route model in the form of a sequence of nodes is used: where si are relay servers, g is the target receiving server [24, 27].
The probability that user u was the source of the request through the confidential communication system (CCS), in the absence of a priori information, is defined as: where PS is the set of potential sources of the request. If the intruder has information about the node through which the traffic passed, the set of possible routes decreases. In the case of building a route without repetitions (without cycles), the number of possible options is calculated as a combination of: γ= 〈 s1 , s2 , … , sn , g 〉 ,
, C bj − 1= j ! ((bb−− 11−) ! j ) !
, N = ( b− 1 ) j, (10) (11) (12) (13) where b is the number of available repeaters, j is the length of the route. If the use of repetitions (loops) is allowed, then the number of possible route options is defined as:
These mathematical models allow for a formalized assessment of the level of user anonymity
and the probability of channel compromise. The functional blocks of the system implement: access
control based on policies and the Zero Trust model [
In a situation where an attacker captures the first node of the route (or another part of the request transmission path), he can partially exclude certain nodes from the set of potential request sources, thereby reducing the degree of user anonymity. In this case, a model with filtering is used to calculate the updated conditional probability that user u initiated the request:
P (b , j , m, k )=
1
| P S|− | NS|
,
(14)
where | P S| is the initial number of possible sources (potential users), | NS | is the number of
excluded (known or compromised) route nodes, b is the route length, j , m, k are route selection
parameters (e.g., depth, bypass strategies, number of mixes). The formula reflects the decrease in
anonymity: the more nodes an attacker can exclude (i.e. | NS | increases), the higher the probability
of correctly identifying the source of the request [
b P ( b )= ∑ P ( b , j , m , k ), (15)
k =0
In the case of an arbitrary access route length in a confidential communication system (CCS),
when all valid lengths have the same probability (uniform distribution), the total probability that
user uuu was the source of the request is calculated by the formula:
(16)
(18)
(19)
adv ( F )= | P ( AF = 1 )− P ( AU = 1 )|,
(17)
where AF is the attacker who has access to the function F , AU is the same attacker, but with
access to a random (ideal) generator U , P(⋅) is the probability of a successful attack [
The level of reliability of the cryptographic function F under the given conditions is defined as: inse c F (t , q )= adv ( F ), where A( t , q ) is the set of possible attacks that have no more than t computational steps and q requests. The formalized value indicates the highest probability of successful hacking of the function F under conditions of limited attacker resources. The lower the value of insec F , the higher the cryptographic strength of the function in a particular environment.
In the context of a cryptographic protocol E : K × R × P →C ,describing a probabilistic
encryption scheme, the security of the system against a Chosen Plaintext Attack (CPA) is evaluated
using the expression [
P ( b )=
1 b∑max ∑j P ( j , m , k ),
b max−b min+1 j =bmin k =0
where b min and b max are minimum and maximum route length, j is the specific route length,k
is the route parameter, such as the number of known nodes or the level of filtering, P ( j , m, k ) is
the conditional probability with fixed route parameters. The model takes into account the
variability of the data transmission route in a secure system and allows you to assess the degree of
user anonymity. The uniform distribution of lengths means that all possible routing scenarios are
considered equally likely, which is a typical assumption in systems such as Mix-net or Tor [
adv CPA ( E )=| P ( A Ek =1 )− P ( AU =1 ) |, where A is the attack algorithm (attacker), E k is the encryption with a secret key k, U is the random function (the standard of complete unpredictability). The indicator reflects the difference between the probability that an attacker will successfully distinguish encryption with a real algorithm from a random function. The lower the value adv CPA, the more resistant the system is to CPA attacks, meaning that an attacker cannot effectively distinguish between ciphertexts, even if he can choose plaintexts.
Accordingly, a system is considered CPA-resistant if:
insec CEPA ( t , q , l )=adv CPA ( E )<ϵ ,
(20)
where ϵ is an acceptable level of cryptographic resistance. This formalized approach provides a
basis for building a software system that guarantees a high level of confidential access and can
adapt to modern information security requirements in computer networks. It combines theoretical
principles with practical cryptographic tools, allowing the implementation of attack-resistant
architectures with support for dynamic access policy, SIEM integration [
To extend the formalized approach to building a software architecture for confidential access in computer networks, mathematical models are used that take into account time dynamics, multilevel trust, adaptability of access policies, and attack resistance. They are truly integrated with Zero Trust mechanisms and analytical capabilities of SIEM systems, creating a flexible and secure information processing platform. Dynamic privacy assurance feature:
Aconf ( s , o , t )= α ∙ Access ( s , o )∙ e− λt,
The formula estimates the level of confidential access of a subject s to an object o at the moment of time t, where α is the weighting factor of the resource’s criticality, λ is the threat intensity, Access ( s , o ) is the access permission (1 or 0). The exponential decay reflects the loss of trust in the resource over time.
Effective level of access of an entity: where f ( u , t ) is the user behavior function u, β is the sensitivity of the SIEM algorithm. This model takes into account changes in behavior to predict potential threats. Probability of anomaly detection P anomaly ( u , t ) assesses how likely the system is to detect suspicious user behavior u at the time t, taking into account the intensity of deviations f ( u , t ) and sensitivity of the SIEM mechanism β .
Assessment of user confidence [
Q trust ( s )= ω 1 ∙ Aut h ( s )+ ω 2 ∙ Hist ( s )+ ω 3 ∙ SIEM ( s ), (25) τ policy ( t )= dC ( t ) =− λ ∙ e−λt,
dt
P anomaly ( u , t )=1−e−β ∙ f ( u ,t ), where δ ( s , o )=1, if access is allowed, and 0 otherwise, g ( s , o ) is the access rank. The formula summarizes the subject’s access levels to all allowed objects, determining the subject’s real influence in the system.
Privacy loss rate:
The rate of privacy loss τ policy ( t ) shows how quickly the probability of maintaining data confidentiality decreases over time. This indicator is used to dynamically adjust access policies in accordance with the rate of information security loss.
Probability of anomaly detection [
Privacy gradient in parameter space:
D adj ( p , t )=∇ p C ( p , t ), (26)
Privacy gradient in the parameter space D adj ( p , t ) reflects how the level of confidentialityC of a resource changes at a certain point in time t depending on its characteristics. This indicator allows you to determine in which direction you need to change access or security parameters to ensure a stable level of confidentiality. It is a useful tool for dynamically adjusting security policies, especially in a changing environment or at increased risk.
System resistance to crypto attacks [
The system’s resistance to cryptoattacks ζ resilience determines the worst-case security scenario by comparing the vulnerability to general probabilistic attacks insec F ( t , q ) and chosen plaintext attacks (CPA) insec CEPA ( t , q , l ). The minimum value among them is selected, which indicates the least resistant component. The lower this indicator is, the higher the cryptographic reliability of the system as a whole. It allows you to formally evaluate the effectiveness of encryption algorithms and justify their suitability for use in a secure access architecture.
Average effective privilege: (28) (29) 1 n
∑ Access ( s i , oi ) ∙ r ( oi ), Φ access= n i=1
Average effective privilege Φ access allows you to quantify the overall level of access in the system, taking into account both the fact of access granted and the sensitivity of each object r ( oi ). The formula calculates the average value of privileges for all active access sessions, where each contribution is weighted according to the level of confidentiality of the resource. A high value of the indicator may indicate the risk of excessive access to sensitive objects, which requires increased control.
Privacy entropy is a formalized metric that allows to assess the uniformity of privacy distribution among information objects in a computer network. It is determined using the Shannon formula:
❑ H conf = − ∑ ❑ P ( oi ) ∙ lo g2 P ( oi ),
i
where P ( oi ) is the probability that an object contains or processes confidential information.
This indicator allows you to quantify how well the confidential resources are distributed within the
network. If the entropy is high, this indicates an even distribution of confidentiality, when no
single object concentrates a significant amount of critical information. On the other hand, a low
entropy value means that there are objects with an excessive concentration of confidential data,
and these objects are critically vulnerable to attacks or information leaks. This situation requires an
immediate review of security policies in order to redistribute the load, strengthen control, or isolate
the most risky objects. Thus, the privacy entropy indicator plays an important role in making
decisions on adaptive risk management, developing security zones, and determining priority areas
for the application of cryptographic protection and monitoring [
Thus, the formula allows to implement an adaptive security policy that automatically adjusts
according to the current user behavior and trust assessment, which is a key element in the
implementation of a modern architecture of confidential access to computer networks [
According to contemporary digital-security requirements, the architecture of a software system
for confidential access to the information resources of computer networks must provide flexibility,
scalability, and resistance to a wide range of attacks [
The administrative console enables administrators to manage access policies, configure
modules, and monitor incidents in real time. It communicates with other system components—
including security controllers, SIEM modules, entry points, and traffic repeaters—to ensure
operational management and rapid incident response. In emergency situations, alternative
communication channels allow direct control of critical infrastructure elements without the need
for intermediate services. Information probes continuously monitor user actions and suspicious
requests, transmitting these data to SIEM systems for analysis and event correlation [
The architecture devotes special attention to organising confidential access to data repositories.
Requests to such repositories are routed through isolated paths created by virtual private networks
(VPNs) that employ tunnelling cryptography [
To verify the reliability of the architecture, a formalised approach grounded in cryptanalytic
models is employed to estimate the probability of information disclosure, privacy degradation over
time, and the effectiveness of the implemented protocols. The system incorporates mathematical
mechanisms for access ranking, dynamic risk analysis, and assessment of resistance to plaintext
attacks, thereby confirming its high level of cryptographic robustness [
Figure 5 presents an extended sequence diagram of the system that protects confidential access to information resources in computer networks, illustrating a typical interaction scenario among key components while reflecting the principles of Zero Trust, multi-factor authentication (MFA), user-behaviour analysis, and cryptographic traffic protection. The process begins when a user submits an access request through an entry point that enforces MFA and Zero Trust verification. After authentication, behaviour analysis is performed, generating a risk score that is forwarded to the Access Controller. The Access Controller conducts contextual authorisation, evaluates applicable access policies, and then issues a decision to allow or deny access. All actions are logged in the SIEM, and a denial automatically triggers an incident notification for administrators. If access is granted, a secure tunnel (WireGuard/IPsec) is established, through which data are exchanged with the information resource (Storage/API). These data remain encrypted during transit, are decrypted on the user’s side, and all access events are recorded for subsequent audit. The diagram thus clearly depicts the phased execution of a secure-access scenario for confidential information, encompassing dynamic access policy enforcement, anomaly response, communication-channel protection, and centralised monitoring. The proposed architecture is suitable for scaling and adaptation both in local area networks and in complex distributed infrastructures using cloud solutions. It allows implementing an integrated security system that maintains the confidentiality, integrity and availability of data at all levels of operation, ensuring resistance to modern threats and compliance with international information security standards.
The study has developed an architecture for a software system that enables confidential access to the information resources of computer networks while addressing current digital-security challenges and supporting effective interaction in highly dynamic network environments. The proposed system combines the concepts of Zero Trust, multilevel authorization, cryptographic traffic protection, and the analytical capabilities of SIEM solutions for anomaly detection and incident response. A formalized access-control model based on security ranking has been implemented, allowing the quantitative assessment of information-object security levels. The integrated mathematical mechanisms let security policies adapt in real time, taking into account risk-oriented approaches and user-behaviour factors. The incorporation of modern cryptographic protocols—TLS 1.3, IPsec, and WireGuard—guarantees resistance to transport-layer attacks, while support for multifactor authentication mechanisms ensures a high level of user trust.
Particular attention has been devoted to implementing anonymous access routes to information resources through Mix-net technology and multilevel encryption mechanisms, thereby minimizing the risk of traffic de-anonymization. Owing to its flexible, modular structure, the system can be scaled to a variety of deployment scenarios—ranging from local networks to cloud platforms— while supporting open interfaces and external software components.
The proposed architecture not only formalizes secure-access processes but also creates an adaptive environment for executing information-security strategies in real time. The system’s practical value lies in its capability to integrate with existing information infrastructures and satisfy industry standards (ISO/IEC 27001, NIST SP 800-207, etc.), making it suitable for protecting critical information assets in the public, financial, educational, and medical sectors.
Thus, the study’s results indicate that the developed architecture for the confidential-access software system provides a high level of information security, robust resistance to contemporary cyber threats, and the capacity to evolve further as digital-environment risks continue to develop.
While preparing this work, the authors used the AI programs Grammarly Pro to correct text grammar and Strike Plagiarism to search for possible plagiarism. After using this tool, the authors reviewed and edited the content as needed and took full responsibility for the publication’s content.