Authentication device with sentiment analysis for secure access control

Abstract

Authentication device with sentiment analysis for secure access control, comprising a modular computer-based structure with coordinated subsystems for data acquisition, affective analysis, behavioral modeling, and access decision computation, wherein the device acquires multimodal input data from a user via an input module that captures text information, speech signals, and facial images, wherein each data stream is encrypted, time-stamped, and synchronized for analysis, and wherein a sentiment analysis unit processes the acquired data by applying speech processing to interpret lexical and syntactic structures, acoustic signal processing to evaluate tonal and prosodic features, and image processing to detect facial expressions and micro-movements in order to generate a sentiment vector representing the user's affective state.

Description

Technical field

[0001] The present invention belongs to the field of computer security and artificial intelligence. In particular, it relates to an authentication device with sentiment analysis, designed to improve secure access control by integrating affective computing and behavior-based analytics into the user verification process. The invention lies within the field of computational intelligence for identity validation and offers a multi-layered, adaptive, and emotion-sensitive mechanism for authenticating authorized users in computer and communication environments. State of the art

[0002] Authentication is the central element of information security and traditionally relies on static access data such as passwords, personal identification numbers, or conventional biometric features. Although these methods are widely used, they are increasingly proving inadequate against modern threats such as social engineering attacks, phishing, spoofing, or data leaks of access credentials. Biometric systems, while more secure than password-based methods, can be fooled by high-resolution replicas or synthetic images generated using modern deepfake technologies.

[0003] Behavioral authentication methods attempt to mitigate such weaknesses by analyzing user activity patterns such as keystrokes or mouse movements. However, these methods evaluate only mechanical characteristics and fail to capture psychological or emotional signals that reflect the user's actual intent or the situational context.

[0004] Sentiment analysis—the computer-aided evaluation of emotional states based on text, speech, or facial expressions—has primarily been used in fields such as marketing, opinion research, and human-computer interaction. Its potential for secure authentication has not yet been realized because existing access control systems rely solely on binary identity matching and do not perform adaptive behavioral assessment. They offer no way to evaluate a user's affective consistency or to detect coercion, stress, or identity impersonation after successful login. Therefore, access control remains vulnerable to subtle but serious forms of abuse where the login credentials are genuine, but the user's intent has been manipulated.

[0005] There is therefore an urgent need for an intelligent authentication device that combines cognitive and emotional indicators with conventional identity parameters. Such a system should be able to recognize genuine behavioral sentiment patterns, assess the consistency between linguistic, acoustic, and visual modalities, and dynamically adapt to legitimate emotional changes, while simultaneously maintaining high robustness against spoofing or forced access attempts. Description of the invention

[0006] The present invention relates to an authentication device with sentiment analysis designed to assess, verify, and continuously monitor the emotional and behavioral integrity of a user requesting access to a protected computing resource. The device integrates multimodal sentiment detection, encompassing textual, acoustic, and visual inputs. Each of these inputs is analyzed by specialized artificial intelligence modules to generate a unified emotional signature. This signature is compared to a behavioral database established during previously authenticated sessions. A trust-score algorithm calculates the degree of congruence between the current sentiment vector and the stored database, with the calculated value determining whether access is granted, restricted, or denied.

[0007] The invention introduces an adaptive security mechanism in which deviations in the affective pattern trigger additional verifications or graded restrictions of user rights. Furthermore, the device performs continuous verification during an active session, ensuring that authorization is maintained only as long as affective coherence persists. Through this real-time monitoring of the emotional and behavioral state, the invention offers enhanced protection against identity theft, coercion, and unauthorized disclosure of access data. The device's architecture allows for seamless integration into existing authentication frameworks and provides self-learning functionality to account for the user's natural emotional development without compromising system security. Detailed description of the invention

[0008] The invention is realized as a modular, computer-aided device comprising coordinated subsystems for data acquisition, affective analysis, behavioral modeling, and access decision calculation. During operation, the device receives multimodal user input via an input acquisition module that captures text information, speech signals, and facial images. Each data stream is encrypted, time-stamped, and synchronized for further analysis. The sentiment analysis unit applies natural language processing techniques to interpret lexical and syntactic structures, uses acoustic signal processing to evaluate tonal and prosodic features, and employs machine vision algorithms to recognize facial expressions and micro-movements.These operations provide quantitative values ​​for emotional polarity, intensity, and stability, which together form a sentiment vector representing the user's current affective state.

[0009] A behavioral base generator stores an affective reference profile for each authorized user, derived from previously verified access sessions. This base reflects characteristic emotional distributions observed during normal authentications and interactions. The reference profile is stored in encrypted memory and regularly updated through reinforcement learning, enabling the device to distinguish natural emotional fluctuations from anomalous behavioral patterns.

[0010] The access decision control continuously compares the current sentiment vector with the stored baseline and calculates a confidence score based on statistical similarity. If the calculated value exceeds a predefined threshold, the user is authenticated or the session validity is maintained. If the value is below this threshold, the device triggers an adaptive security measure, which may include additional authentication, a temporary session suspension, or a restriction of user rights. In this way, the device combines psychological context with technical verification, enabling particularly robust access control.

[0011] In practical implementation, the invention functions as software running on a secure microprocessor or in a distributed cloud computing environment. The device can operate as a standalone module or be integrated as a component into existing enterprise authentication systems. It communicates with conventional password or biometric systems via standardized programming interfaces, extending them with a layer of emotional intelligence without altering the existing infrastructure. All collected data is processed in accordance with applicable data protection and security regulations, ensuring that the affective analysis is used exclusively for legitimate security purposes.

[0012] The device operates completely autonomously and requires no human supervision. During each login process, multimodal input is captured, sentiment features are extracted, and then compared to the behavioral baseline. If the calculated match falls within an acceptable range, access is granted; otherwise, the system requests additional verification or denies access. After successful authentication, the device continuously monitors emotional coherence in real time. Sudden or sustained deviations from established affective patterns trigger predefined protective measures.

[0013] Through this closed-loop feedback process, the system continuously learns from real-world interactions and improves the accuracy of future decisions. The invention thus achieves secure, adaptive, and psychologically sound authentication by combining sentiment analysis, behavioral modeling, and reinforcement learning within a unified, computer-based framework. This ensures that access control is not only identity-specific but also intention-aligned, significantly reducing the risks of data misuse, identity fraud, or emotional manipulation.

Claims

[1] Authentication device with sentiment analysis for secure access control, comprising a modular computer-based structure with coordinated subsystems for data acquisition, affective analysis, behavioral modeling and access decision computation, wherein the device acquires multimodal input data from a user via an input module that captures text information, speech signals and facial images, wherein each data stream is encrypted, time-stamped and synchronized for analysis, and wherein a sentiment analysis unit processes the acquired data by applying speech processing to interpret lexical and syntactic structures, acoustic signal processing to evaluate tonal and prosodic features and image processing to detect facial expressions and micro-movements in order to generate a sentiment vector representing the affective state of the user. [2] Device according to claim 1, characterized by , that a behavior base generator is provided which maintains an affective reference profile for each authorized user, derived from verified access sessions, stored in encrypted memory and periodically updated using reinforcement learning algorithms to distinguish natural emotional fluctuations from anomalous behavioral deviations. [3] Device according to any one of the preceding claims, characterized by , that an access decision control unit continuously compares the current sentiment vector with the stored affective reference profile, calculates a confidence score based on statistical similarity, and authorizes or restricts access depending on whether a predefined threshold is exceeded or not met. [4] Device according to any one of the preceding claims, characterized bythat the device is implemented on a secure microprocessor or within a distributed cloud computing environment, operates autonomously as a standalone unit or as an integrated component within a corporate authentication system, and is connected to conventional password or biometric modules via standardized communication protocols, ensuring compliance with data protection regulations. [5] Device according to any one of the preceding claims, characterized bythat the device performs continuous real-time authentication during an active user session, monitors emotional coherence to detect persistent deviations from the baseline affective profile, and, upon detection of an inconsistency, activates a predefined protective measure including secondary verification, temporary blocking, or access restriction, thereby achieving adaptive and psychologically informed secure access control.