Deceptive Denial: Concepts and Applications

• Deceptive denial is a systematic approach that purposefully misleads adversaries by concealing or distorting true information.
• It is employed in domains such as cybersecurity, AI, wireless communications, and organizational transparency to strategically deny access to genuine resources.
• Formal models and algorithms, including Denial Logic, safety games, and deceptive beamforming, validate its effectiveness and guide mitigation strategies.

Deceptive denial is a multifaceted concept referring to the deliberate obstruction, denial, or undermining of access to truth, services, or resources—typically through systematic deception rather than overt denial. It encompasses strategies in artificial intelligence, cyber defense, security games, organizational communication, wireless systems, formal logic, and information attacks, unified by the intent to mislead recipients or adversaries into false negative conclusions. Manifestations range from LLMs prompted to generate denials of factual information, to cyber systems masquerading as unavailable, to agents or organizations withholding genuine capabilities or usage with strategic falsehoods.

1. Conceptual Foundations and Definitions

Deceptive denial denotes the intentional withholding, misrepresentation, or denial of truths, services, or capabilities in a way that induces false beliefs or actions in adversaries, systems, or human observers. Unlike brute-force denial (e.g., physical removal of service), deceptive denial injects ambiguity or synthesizes denial through information manipulation.

Key definitions across domains include:

• Cybersecurity/Deception Technology: Deliberate removal or hiding of genuine assets/services to skew attacker decisions, often using decoys or ambiguity rather than outright blocking (Fraunholz et al., 2018).
• Safety Games: Defender deploys "traps" (hidden real sinks) or "fake targets" (false attractors) so that an attacker is denied access to real targets by misdirection or hidden hazards, constituting denial via deception on a graph hypergame (Kulkarni et al., 2024).
• Organizational Transparency: "Deceptive denial" is identified as the practice of understating (or entirely denying) an organization's true use of AI, particularly when actual use is hidden behind "no AI" public claims (Nyilasy et al., 6 Jul 2025).
• LLMs and Informational Attacks: Instructing a LLM to generate factually false outputs, or to withdraw ("deny") belief in truths when presented with sophisticated, hard-to-falsify deceptive evidence (Long et al., 29 Jul 2025, Wan et al., 9 Jan 2026).
• Formal Logic: Denial Logic (DL) formalizes agents whose justified beliefs are always false—i.e., every "justified" assertion is a denial of truth (Lengyel et al., 2012).
• Wireless Communications: Transmission of deceptive signals or beamforming to mislead eavesdroppers, thereby denying them correct channel, range, or Doppler information (Chrysanidis et al., 6 Mar 2025).
• Information-Induced DoS in ITS: Delivery of crafted misinformation to induce human/system-level denial of service or safe operation in intelligent transportation systems (Yang et al., 2024).

A common property is the dual action: the system not only refuses access/truth, but does so in a manner that leaves the target misinformed about the true state of affairs.

2. Theoretical and Formal Models

Mathematical and algorithmic models have developed to analyze and synthesize deceptive denial across domains:

Deception Technology Taxonomy

Deceptive denial occupies the "denial" pole in taxonomies of deception techniques—distinguished from deceit, misinformation, camouflage, and obfuscation (Fraunholz et al., 2018). It is modeled as a two-player game, with defender and attacker payoffs formulated according to the success of the deception and associated costs.

Safety Games on Graphs

Hypergames capture the mismatch between a defender's true transition graph and the attacker's perceptual game (which is distorted by deceptive denial means). The defender's optimization problem—placement of traps ("hide the real") and fake targets ("reveal the fiction")—is formalized as a combinatorial optimization over the defender's deceptive winning region, which is shown to be monotone and (sub/super)modular, admitting efficient greedy approximation (Kulkarni et al., 2024).

Denial Logic (DL)

Denial Logic provides a formal system where every justified belief is false, using axiomatics such as $t\!:\!P \rightarrow \neg P$. The addition of negative constant specifications permits modeling agents that systematically deny truths but maintain internal logical coherence (Lengyel et al., 2012).

Information-Induced DoS (IDoS) Games

In ITS, Stackelberg games formalize an attacker (leader) injecting fake demand or misinformation to force denial or disruption, and a defender (follower) recalculating equilibria or control actions. Trust constraints and adaptive learning further harden the system against continuous or adaptive deceptive denial (Yang et al., 2024).

3. Algorithms, Detection, and Implementation Architectures

Sparse Autoencoders and Probing in LLMs

SAEs reveal that deceptive instructions to LLMs shift internal representations along specific, sparse mid-layer features, flipping truthfulness signals and enabling denials of fact without erasing underlying knowledge. Monitoring and patching these features can detect and counteract deceptive denial at inference time (Long et al., 29 Jul 2025).

Misery Digraphs in Cloud Networks

Cloud-based "misery digraph" architectures instantiate deceptive denial by interposing a k-ary tree of decoy nodes between entry points and targets, exponentially increasing the search space for remote attackers and dynamically reconfiguring to invalidate reconnaissance. Legitimate users incur modest performance penalties, whereas attackers' search costs grow exponentially in the depth of the misleading structure (Qasem et al., 2020).

Power-Efficient Deceptive Beamforming

In wireless systems, beamforming can be optimized—via convex quadratic programming and heuristic algorithms—to inject controlled range or Doppler misinformation toward eavesdroppers, achieving denial with lower transmit power than brute-force nulling, and without distorting legitimate communications (Chrysanidis et al., 6 Mar 2025).

Governance for Deceptive Evidence in LLMs

Deceptive Intent Shielding (DIS) implements an analyst module to classify evidence intent ("direct support," "indirect support," "opposition"), appending warnings to potentially deceptive evidence before consumption by decision agents—reducing belief shifts in LLMs exposed to refined false evidence (Wan et al., 9 Jan 2026).

4. Empirical Effects and Quantitative Findings

Psychological and Social Impact

Within organizational behavior, deceptive denial (e.g., denying real AI usage) produces a substantial negative impact on moral judgments (ΔAttitude = –1.27 points, d=0.97) and purchase intentions (ΔPI = –1.11, d=0.78), mediated fully by perceived betrayal (Indirect ≈ –0.32, 95% CI [–0.46, –0.20]) (Nyilasy et al., 6 Jul 2025).

LLM Susceptibility and Control

When exposed to sophisticated, hard-to-falsify deceptive evidence, belief scores in leading LLMs rise by an average of 93.0%, fundamentally shifting model recommendations (e.g., 29% of GPT-5 responses flip from cautious to risky). Deceptive Intent Shielding reduces belief shifts by ~30% and restores binary accuracy by up to 288.8% (Wan et al., 9 Jan 2026).

Cyber Defense Effectiveness

Honeypot-based denial traps and misery digraphs show >95% detection of malicious scans with minimal (<0.1%) false positive rates, and introduce time-to-detection (TTD) and search cost penalties to attackers while maintaining operational efficiency for legitimate users (Fraunholz et al., 2018, Qasem et al., 2020).

Wireless Denial Efficiency

Power-efficient deceptive beamforming outperforms null-forming in transmit power (up to 120% less) while achieving complete spoofing of eavesdropper range–Doppler estimates and maintaining mainlobe directionality (Chrysanidis et al., 6 Mar 2025).

Game-Theoretic ITS Defense

Strategically injected fake demands cause targeted impact (TI) and network impact (NI) up to double those of random attacks in navigational systems; trust constraints and learning mechanisms limit attack efficacy, with provable regret and stability bounds (Yang et al., 2024).

5. Ethical, Legal, and Policy Considerations

Entrapment, Liability, and Compliance

Deployers of deceptive denial must avoid entrapment (non-law enforcement contexts), respect privacy boundaries (avoid capturing personal data), and ensure legal compliance (e.g., Wiretap Act, DMCA, CFAA) when routing or disrupting flows (Fraunholz et al., 2018).

Organizational Transparency and Regulatory Guidance

Deceptive denial erodes stakeholder trust, with pronounced effects in sensitive domains (insurance, health, banking). Regulatory bodies (e.g., FTC) are urged to expand guidelines to encompass "AI-washing" in both false positives and false negatives—mandating full disclosure of AI usage and supporting AI-literacy initiatives (Nyilasy et al., 6 Jul 2025).

Practical Recommendations

Best practices include blending denial with obfuscation strategies, leveraging cognitive biases in defense, monitoring probe and dwell metrics, and isolating suspicious flows in forensically robust sandboxes. Adaptive, learning-driven denial defenses and multi-layered trust systems are essential in adversarial environments (Fraunholz et al., 2018, Yang et al., 2024).

Open Research Questions

Challenges remain in advancing adaptive denial mechanisms, quantifying psychological impact on attackers and users, hardening denial traps against fingerprinting, and harmonizing cross-jurisdiction legal regimes. The sophistication of deceptive denial in both information and cognitive domains complicates detection and mitigation.

6. Philosophical and Logical Perspectives

Denial Logic formalizes the pattern of agents who systematically "justify" false beliefs, isolating the structure of self-deception and philosophical skepticism. The introduction of negative constant specifications allows the modeling of consistent but invariably false epistemic states. The "Blue Pill" theorem demonstrates that, under coherence, it's possible to construct models where all previously justified falsehoods become true, capturing the epistemic indeterminacy in "deceptive denial" (Lengyel et al., 2012).

Classic philosophical motifs—Descartes’s radical doubt, Putnam's brain in a vat, or climate-change denial—are shown to be structurally instances of deceptive denial, where agents preserve internal coherence at the expense of correlation with external truth.

---

Deceptive denial thus comprises a collection of techniques and phenomena at the intersection of security, epistemology, AI behavior, and organizational conduct, unified by their reliance on robed unavailability or false negative assertion as a defensive or manipulative act. The research corpus demonstrates that detection, mitigation, and ethical governance of such phenomena require tightly integrated technical, cognitive, and policy solutions.