Virtual Radial Keyboard

• Virtual Radial Keyboard is a secure VR text-entry system that arranges 27 keys alphabetically in a circular, color-coded layout for robust security and ease of navigation.
• It employs an angular-based selection mechanism with dynamic key resizing and per-keystroke random rotation to thwart spatial inference attacks.
• Empirical evaluations reveal a trade-off between reduced typing speeds and substantially improved resistance against both basic and advanced keystroke inference attacks.

The Virtual Radial Keyboard is a secure and usability-focused text-entry method developed for virtual reality (VR) environments. Departing from the conventional QWERTY layout, it arranges all keys alphabetically in a circular ring to facilitate angular-based selection via controller rotation. Its defining features include dynamic key resizing for enlarged target acquisition, a randomized per-keystroke rotation mechanism that disrupts fixed spatial mappings, and grouping of letter sectors via color coding. This architecture is explicitly designed to foil keystroke inference attacks while accommodating the ergonomic constraints of VR input devices (Huang et al., 9 Jan 2026).

1. Structural Layout and Selection Mechanism

The keyboard comprises 27 keys (A–Z plus space), arrayed clockwise in a single ring and subdivided into six color-coded groups of four or five letters. Two function keys (Enter, Backspace) reside at the center of the circular configuration. The space key is randomly assigned a position among the alphabetic sectors at the beginning of each session.

Selection is mediated by ray casting, where users rotate their wrist to control a selection ray. When the intersection of the ray and the ring falls within a given sector’s angular bounds, that sector is flagged as “under selection.” This circumvents the need for precise spatial targeting and leverages the natural motion afforded by VR controllers.

2. Dynamic Key Resizing

To further enhance usability, the active sector—whenever a key is under selection—expands radially outward by an increment equal to the nominal key-width, doubling its radial thickness. The mathematical formalism is as follows:

• Number of sectors, $K=27$,
• Nominal angular width per sector, $\theta_0 = \frac{2\pi}{K}$,
• Radial bounds, $r \in [r_{in}, r_{out}]$,
• Upon activation of sector $j$, the outer radius is set to

$r_{out}' = r_{out} + (r_{out} - r_{in})$

This expansion mitigates accidental departures to adjacent sectors, increasing tolerance for coarse aiming and providing a prominent visual indicator of selection. The approach is tailored to the spatial limitations and motion fidelity characteristic of VR environments.

3. Per-Keystroke Non-Intrusive Randomization

A critical departure from standard text-entry paradigms is the randomized rotation mechanism. After each keystroke, the ring rotates by 0 or one sector ($\theta_0$) clockwise, governed by a session-specific bias variable $p_{left} \sim \text{Uniform}(0.2, 0.8)$ reflecting the likelihood of expansion “to the left.” An initial offset $\phi_0$ is randomized at session start. The rotation rule is:

• If the selection ray enters the key’s left half, no rotation; otherwise, a $+\theta_0$ rotation is applied.

This mechanism preserves relative letter order (maintaining alphabet adjacency) but invalidates absolute spatial mappings, thus impeding adversarial attempts to map controller angles to key identities on replay or screen recording. The rotation magnitude never exceeds one sector per keystroke, resulting in a slow, unpredictable drift in keyboard orientation.

4. Empirical Security Evaluation

Two attacker models were examined:

• Basic Attack: Assumes knowledge of initial offset $\phi_0$ and ignores subsequent rotations.
• Advanced Attack: Employs Viterbi beam search (beam size 1000) using a trigram LLM (Norvig's big.txt) to correlated angular transitions.

Security was quantified with Identical Character Ratio (ICR) and Semantic Similarity (SS) metrics:

Method	ICR (Basic)	ICR (Advanced)	SS (Advanced)
QWERTY (none)	~1.00	~1.00	~1.00
ISPR (basic)	~0.85	~0.08	~0.12
Radial (basic)	~0.05	~0.18	~0.10

In the radial keyboard scenario, even advanced attacks plateau at $\sim$17% character accuracy and random semantic similarity, indicating exponential decoding complexity (search space $\sim 754\times 2^n$, $n$ = keystrokes). This represents a substantial increase in security over prior spatially randomized (ISPR) or standard QWERTY approaches.

5. Usability and Performance Metrics

A within-subject study (N=30) compared Radial, ISPR, and QWERTY keyboards:

Keyboard	WPM (Median)	Total Error Rate (%)	SUS Usability	NASA-TLX Workload
QWERTY	11.39	2.4	85	16.5
ISPR	8.03	3.6	66	38
Radial	5.13	3.1	55	47

The Radial keyboard yielded lower entry speeds and higher subjective workload (NASA-TLX) compared to QWERTY, attributable to non-standard alphabetical ordering and dynamic spatial drift. Physical demand was consistently low—users could operate the device with minimal wrist movement. However, cognitive demands were elevated, with several participants expecting substantial performance improvement given longitudinal exposure.

6. Trade-Offs, Design Implications, and Refinement Directions

The primary trade-off is between inference resistance and text-entry efficiency; the randomized dynamic layout substantially increases security but results in approximately a 50% reduction in WPM versus QWERTY. Design recommendations include:

• Expanding the selected sector to triple width on double-clicks and vetoing rotation options that reduce cursor-boundary distance, which may ameliorate accidental mis-selections.
• Implementing visual guide lines or gradient shading to signal infinite outward sector extension.
• Exploring alternative alphabetic groupings (e.g., clusters based on frequent bigram pairs) to increase entry rates.

This suggests a potentially steep learning curve but also the possibility of adaptation and proficiency gains with practice. Future directions proposed include 3D spatial variants, haptic augmentation, enhanced visual cues, and evaluation against deep-learning-based adversarial models. Longitudinal adaptation studies and interface refinements likewise remain open for exploration (Huang et al., 9 Jan 2026).

7. Summary and Position Within the Text-Entry Design Space

The Virtual Radial Keyboard constitutes a new security-oriented point in VR text-entry design, achieving strong inference resistance through combinatorics and per-keystroke layout randomization, while employing dynamic expansion for usability. It requires only standard VR hardware (controller and headset) and maintains a coherent, learnable alphabetic order. While performance in raw speed and usability currently lags established QWERTY conventions, qualitative feedback and system flexibility indicate promising avenues for iterative improvement and increased adoption in security-critical VR applications.