False Reality: Fabrications & Perceptions

• False Reality is an umbrella concept describing the divergence between appearances and underlying authenticity, often via fabricated narratives and synthetic media.
• It encompasses diverse domains such as disinformation, sensor-manipulated VR, AI-generated content, and philosophical debates on realism.
• Researchers face significant challenges in detection, attribution, and establishing reliable authenticity in increasingly deceptive information landscapes.

False reality denotes a family of problems in which what is taken as real is displaced by fabricated narratives, synthetic media, manipulated sensor streams, or theory-lidden appearances whose relation to underlying events is contested. In contemporary technical literature, the expression spans disinformation and fake news, AI-generated media and media authentication, sensor-induced deception in virtual and mixed reality, and foundational disputes over whether observable phenomena transparently reveal what exists (Cardenuto et al., 2023, Jiang et al., 11 Aug 2025, Bertlmann, 2020).

1. Scope and principal meanings

The literature does not use false reality as a single settled term. This suggests an umbrella concept whose common feature is not simple falsity, but a breakdown between appearance, representation, and underlying state. In media studies, the issue is the construction of misleading public narratives; in AI forensics, it is the generation of plausible but synthetic artifacts; in cyber-physical systems, it is the corruption of sensed reality; and in philosophy of physics, it is the possibility that observable or calculational structures should not be identified with ontology (Cardenuto et al., 2023, Jiang et al., 11 Aug 2025, Plotnitsky et al., 2015).

Domain	Core mechanism	Representative arXiv ids
Disinformation and fake news	Propagation of intentionally false or misleading narratives	(Mukhaini et al., 2022, Espinoza et al., 2021, Růžička et al., 2019)
Synthetic media and forensics	AI-generated or AI-augmented images, video, audio, text, and immersive media	(Cardenuto et al., 2023, Sun et al., 2024, Radharapu et al., 2024)
VR, XR, and cyber-physical systems	Manipulated sensor measurements and false data injection	(Jiang et al., 11 Aug 2025, Pant et al., 2024, Sisini et al., 2023)
Quantum and philosophical discourse	Tension between reality, realism, truth values, and representation	(Bertlmann, 2020, Boughn, 2019, Bolotin, 2020)

A recurring distinction across these areas is between a system’s internal or public-facing representation and the event, process, or ontology it purports to track. False reality, in this broad sense, is not restricted to fabricated content; it also includes fabricated authenticity judgments, fabricated sensor worlds, and philosophically misleading reifications of successful formalisms.

2. Public narratives, disinformation, and fake news

In the misinformation literature, an important baseline distinction is between misinformation and disinformation: misinformation is false or misleading information without intent to deceive, whereas disinformation is false or misleading information where intent is present (Espinoza et al., 2021). One line of work models disinformation as a wave-like social phenomenon. In that account, propagation is mapped to $v=\lambda f$, intensity to $I \propto A^2$, skepticism functions as a refractive medium that lowers propagation speed, and polarization functions as an evidence-based filter rather than mere partisan division. The proposed intervention is therefore twofold: skepticism as the default mode, and polarization as a filter mechanism to analyze claims on the basis of evidence (Espinoza et al., 2021).

A more explicitly cultural account treats fake news as a contemporary myth-form rather than only as misinformation or propaganda. “The Myths of Our Time: Fake News” frames fake news as something akin to myths, folk tales, and urban legends, and technically realizes that claim through a generation pipeline built from a corpus of 245,973 articles and 196,952,689 total words, topic-specific subsets, word-level LSTM models, beam search, Levenshtein-distance filtering, and a blog presentation using a fictional journalist persona, “Misun Lean,” whose portrait is generated with PGGAN (Růžička et al., 2019). The project’s importance lies in showing that false reality is infrastructural: it depends not only on text generation, but also on tags, images, bylines, and interface conventions that stabilize credibility.

A more conventional detection study is “FALSE: Fake News Automatic and Lightweight Solution,” which uses the LIAR benchmark of 128,000 human-labeled short statements from PolitiFact.com, with 14 attributes and six standard truthfulness categories: pants-fire, false, barely-true, half-true, mostly-true, and true (Mukhaini et al., 2022). The study emphasizes metadata and the five historical truthfulness count features—barely true counts, false counts, half true counts, mostly true counts, and pants on fire counts—rather than explicit text representation. It reports high correlations such as 0.9136604 between barely true counts and half true counts, and classifier performance that reaches 99.9% accuracy for MCR, JRip, J48, TRF, and TRT. At the same time, the paper does not report a train/test split, cross-validation design, random seed, class balancing method, hyperparameter settings, or statistical significance tests, and it appears to rely heavily on speaker truth-history counters. The resulting false-reality problem is therefore double: fabricated political claims on the input side, and potentially overconfident authenticity judgments on the analytic side.

3. Synthetic media, deepfakes, and authenticity infrastructures

The broadest forensic framing appears in “The Age of Synthetic Realities,” which defines synthetic realities as “any contextual digital creation or augmentation enabled by artificial intelligence methods,” regardless of the intent to deceive (Cardenuto et al., 2023). The category includes synthetic images, videos, audio, text, NeRFs, metaverse environments, and deepfakes as a cross-modal subtype centered on identity or trait replacement. The paper stresses that the core forensic tasks are detection, attribution, explainability, provenance, and multimodal contextual fusion. It also argues that the distinction between reality and fiction is “getting thinner daily,” particularly when synthetic media are embedded in fake news, disinformation, and misinformation workflows.

Deepfake detection itself can become a site of false reality. “Real is not True: Backdoor Attacks Against Deepfake Detection” introduces Bad-Deepfake, a training-time poisoning attack against deepfake detectors in which a small poisoned subset induces a hidden rule: fake media carrying an optimized trigger are classified as “Real” (Sun et al., 2024). The attack uses an $L_\infty$-constrained additive trigger with $\epsilon = 2/255$, projected gradient descent for $T=50$ steps, and a Filtering and Updating Strategy for poisoned sample selection. On FaceForensics++, with SE-ResNeXt as the documented detector, the abstract reports a 100% attack success rate, while benign accuracy remains comparable to the clean model. The epistemic consequence is central: once the detector is compromised, a “Real” label no longer means authentic content, only content matching the attacker’s trigger condition.

A capture-time alternative is proposed in “RealSeal,” which argues that the field should stop trying to mark all fake content after the fact and instead “affix a certificate of authenticity with only the real images and hop[ing] that fake images can’t acquire one” (Radharapu et al., 2024). RealSeal combines sensing, scoring, and signing inside a secure execution environment. It proposes multisensory capture—3D spatial, thermal, motion, and auditory inputs—an aggregated realism or credibility score, and cryptographic signing of the image-plus-metadata bundle. The proposal also stresses secure boot, TEEs or enclaves, HSM-like key storage, and compatibility with provenance frameworks such as C2PA. Yet the realism score is not formally defined, no prototype or benchmark is reported, and there are no latency, calibration, AUROC, or error-rate measurements. Here false reality is addressed at the infrastructure layer, but the proposed remedy remains largely conceptual.

4. Sensor-mediated false reality in virtual and mixed-reality systems

In XR systems, false reality may be induced without modifying software at all. “False Reality: Uncovering Sensor-induced Human-VR Interaction Vulnerability” formalizes an end-to-end pathway from physical signals to human effects: $$\text{physical signals} \rightarrow \text{sensor measurements} \rightarrow \text{VR system services} \rightarrow \text{human perception and actions}$$ and models the transfer from attack signal to perception as

$$G(s)=\frac{sense(s)}{attack(s)}=\frac{F_s(s)F_p(s)F_a(s)H_s(s)}{1-F_s(s)F_p(s)F_a(s)H_s(s)H_a(s)}.$$

The paper demonstrates three pathways on five commercial VR devices: trajectory manipulation through ultrasonic IMU disturbance of HMD locating, avatar manipulation through controller IMU disturbance, and dizziness attacks through magnetic interference with Hall-sensor-based IPD adjustment (Jiang et al., 11 Aug 2025). The reported average service-disturbance success rate is 85% for HMD locating and 87.5% for controller tracking; end-to-end studies report an average distance beyond the safety boundary of 0.597 m, an average robotic end-effector displacement of 0.626 m, IPD bias up to 10 mm at 2 A excitation current, and an average dizziness score of 1.67. The proposed defense is a low-cost vibration module with

$V=(1-e^{-|v_z|})V_{max},$

intended to expose anomalies through tactile mismatch rather than pure visual trust.

A related cyber-physical formulation appears in “MIXED-SENSE,” a mixed reality sensor emulation framework for UAV resilience evaluation against false data injection (Pant et al., 2024). The framework combines Gazebo and a Motion Capture system to emulate proprioceptive and exteroceptive sensor measurements in real time, explicitly including GNSS and camera-like sensing, and uses an empirical approach to recreate latency and noise. The concrete demonstration is an emulated GNSS attack on an actual UAV together with validation of a mitigation strategy using a distributed camera network. In this case, false reality is not an audiovisual illusion for a human subject, but a falsified sensor world fed into a real estimator-and-controller loop.

Theoretical work on VR generalizes this logic. “Theoretical limits of Virtual Reality” models the total physical/computational system by degrees of freedom $\mathbf{C}$, observable outputs by $\mathbf{O}\subseteq \mathbf{C}$, and hidden variables by $I \propto A^2$0 (Sisini et al., 2023). On this account, a virtual world is experientially real for the observer whenever the observable trajectory admits a coherent lawful reconstruction, and a “non-real” or “magical” world becomes possible precisely when hidden interacting degrees of freedom are withheld. The paper’s key limit is that complete deception depends on asymmetry of access: passive observation or restricted interaction can sustain false reality, while sufficiently rich intervention can expose the hidden structure.

5. Philosophical and physical formulations

In quantum foundations, false reality is not usually a term of art, but the underlying problem is the same: whether observable or calculational structures should be taken as literal reality. Bertlmann’s recollection of John Bell presents realism as the “natural position for a scientist,” built around beables rather than observables, and ties that realism to Bell inequalities and the conclusion that local realistic theories are incompatible with Nature (Bertlmann, 2020). At the same time, the paper contrasts this with the information-theoretic line associated with Brukner and Zeilinger, on which the quantum state is “just the mathematical representation of our knowledge about the system,” and with the ambiguous status of virtual particles, which are indispensable to calculation yet not straightforwardly observable. The resulting lesson is not that reality is unreal, but that classical local realism, textbook “measurement” talk, and easy reification of formal entities are all suspect.

“Reality without Realism” makes that distinction explicit by separating reality from realism: reality is that which exists, while realism is our ability to represent, describe, or conceive its independent architecture (Plotnitsky et al., 2015). Plotnitsky’s statistical Copenhagen interpretation treats quantum objects or processes as beyond description, and perhaps beyond conception, while Khrennikov’s PCSFT posits a deeper field-like ontology in which the density operator is interpreted as a covariance operator of a prequantum random field and detector clicks arise from threshold detection. The addendum on the PBR theorem disputes a different route to wavefunction realism by arguing that PBR tacitly assumes a false transfer of ensemble probabilities to individuals with fixed ontic states (Cabbolet, 2024). By contrast, Boughn argues that the very term “reality” “should be avoided at all costs” in physics discourse because it leads into “rabbit hole” debates over collapse, hidden variables, non-locality, and many worlds (Boughn, 2019).

A distinct but related line relocates false reality to perception itself. “Do We Perceive Reality?” presents Donald Hoffman’s thesis that perception tracks fitness, not truth, and reproduces the “Fitness-Beats-Truth Theorem,” according to which “fitness drives truth to extinction with probability at least $I \propto A^2$1” (Klasios, 2022). The same paper brings black holes, the holographic principle, duality, quantum gravity, and special relativity into the argument that spacetime, objects, and temporal flow may be non-fundamental. Maniatis’s “Illusions – a model of mind” similarly argues that all mental processes are “unfree and passive,” that perception is a “bunch” of associations rather than an inner display, and that consciousness arises from “silent communication” in first-person form (Maniatis, 2017). Whitworth’s “The Virtual Reality Conjecture” pushes the strongest ontological version: the physical world is “the digital output of non-physical quantum processing,” so the physical universe is real-as-experienced but not objectively self-existing (Whitworth, 2011).

The question of truth values is made explicit in “Wave-particle duality and the objectiveness of ‘true’ and ‘false’,” which argues that the textbook reading of the double-slit experiment threatens the objectivity of truth itself unless one rejects universal bivalence for slit-specific propositions (Bolotin, 2020). The paper retains only the values true and false, but replaces total valuation by partial bivaluation, allowing truth-value gaps for propositions such as “the quantum particle passes through slit 1” while preserving the objective truth of the exclusive disjunction that it passes through one slit or the other, but not both. False reality, on this reading, arises not because reality is contradictory, but because semantic expectations are too coarse for quantum phenomena.

6. Limits, controversies, and open problems

A common theme across the literature is that false reality is easier to generate than to validate away. In fake-news detection, near-perfect reported metrics can rest on speaker-history shortcuts, missing validation protocols, or domain-specific datasets of short political statements rather than general misinformation environments (Mukhaini et al., 2022). In synthetic media forensics, the field faces open-set recognition, unseen generators, post-processing, mixed-authorship content, and multimodal inconsistency, while the authors of “The Age of Synthetic Realities” emphasize detection, attribution, explainability, and context as unresolved research fronts (Cardenuto et al., 2023). In deepfake detection, the verifier itself can be backdoored, which means authenticity labels may become attacker-controlled rather than truth-tracking (Sun et al., 2024).

Another recurring issue is that proposed defenses often shift, rather than eliminate, the burden of trust. RealSeal moves trust to capture-time scoring, secure hardware, and signed metadata, but it does not yet specify the realism score mathematically or benchmark it operationally (Radharapu et al., 2024). VR defenses based on perceptual fusion improve anomaly noticeability, yet the underlying attacks remain device-specific rather than theoretically closed (Jiang et al., 11 Aug 2025). Mixed-reality UAV testing improves safety and repeatability for attack evaluation, but it studies the effect of deceptive sensing on the estimation-control loop rather than fully reproducing every physical attack chain (Pant et al., 2024). A plausible implication is that robust defense will require not a single detector or credentialing layer, but combinations of provenance, sensor diversity, poisoning-aware training, anomaly detection, and trusted independent validation.

The philosophical literature adds a further complication: there is no consensus on whether false reality is principally a problem of deception, perception, semantics, ontology, or language. Bell-style realism, information-theoretic nonrealism, anti-“reality” pragmatism, interface theories of perception, and virtual-world conjectures all preserve some distinction between appearance and what lies behind it, but they sharply disagree on whether that distinction is representable, measurable, or even meaningful (Bertlmann, 2020, Boughn, 2019). The encyclopedic significance of false reality therefore lies less in a single doctrine than in a structural pattern: a representation, judgment, or experience functions as reality while its grounding conditions are displaced, hidden, poisoned, or philosophically underdetermined.