cs.ROFeb 26, 2026
SaveCybersecurity of Teleoperated Quadruped Robots: A Systematic Survey of Vulnerabilities, Threats, and Open Defense Gaps
Organizations: Department of Informatics, Bioengineering, Robotics and Systems Engineering (DIBRIS), University of Genoa, Genoa, Italy
Abstract
Teleoperated quadruped robots are increasingly deployed in safety-critical missions -- industrial inspection, military reconnaissance, and emergency response -- yet the security of their communication and control infrastructure remains insufficiently characterized. Quadrupeds present distinct security challenges arising from dynamic stability constraints, gait-dependent vulnerability windows, substantial kinetic energy, and elevated operator cognitive load. This survey synthesizes peer-reviewed literature and vulnerability disclosures (2019--2025) to provide comprehensive analysis of cybersecurity threats, consequences, and countermeasures for teleoperated quadruped systems. We contribute: (i) a six-layer attack taxonomy spanning perception manipulation, VR/AR operator targeting, communication disruption, control signal attacks, localization spoofing, and network intrusion; (ii) systematic attack-to-consequence mapping with timing characterization; (iii) Technology Readiness Level classification exposing critical maturity gaps between field-deployed communication protections (TRL 7--9) and experimental perception/operator-layer defenses (TRL 3--5); (iv) comparative security analysis of six commercial platforms; (v) pragmatic deployment guidance stratified by implementation timeline; and (vi) eight prioritized research gaps with implementation roadmaps. Limitations: Platform assessments rely on publicly available information. Attack success rates derive from cited studies under controlled conditions and require domain-specific validation.
Figures & tables
Fig. 1 : Conceptual illustration of communication channel attacks in quadruped robot teleoperation. An adversary can target the wireless link between the operator and robot to intercept, manipulate, or disrupt command and feedback transmissions.
Fig. 2 : Representative commercial quadruped platforms analyzed in this survey. These platforms exhibit diverse security architectures: Spot employs per-device cryptographic keying, Unitree platforms have exhibited fleet-wide credential vulnerabilities, and ANYmal targets industrial inspection with enterprise security requirements. (a) Boston Dynamics Spot; (b) Unitree B2; (c) ANYbotics ANYmal.
Fig. 3 : PRISMA-style flow diagram illustrating the systematic literature selection process. Preliminary database search across IEEE Xplore, ACM Digital Library, Scopus, Web of Science, and arXiv yielded 847 potentially relevant publications. Supplementary sources (n=63) include vendor security documentation, CVE databases, and publications identified through forward/backward citation tracking.
Fig. 4 : Quadruped teleoperation architecture with six-layer attack surface mapping. Red dashed boxes indicate attack vectors targeting each trust boundary zone. The three trust boundaries (Operator, Communication, Robot) correspond to the adversary model’s attack access points. Each crossing between zones represents an exploitable interface where attacks from Layers 1–6 can intercept, manipulate, or disrupt the bidirectional command-feedback pipeline.
| Solution Domain | Approach | Quadruped Limitation |
|---|---|---|
| UAV Security | GPS-centric localization protection | Quadrupeds often operate GPS-denied (indoors, urban canyons) |
| UAV Security | Hover-and-wait on communication loss | Quadrupeds cannot hover; must maintain active balance |
| Industrial Robot | Safety caging and exclusion zones | Quadrupeds operate in unstructured environments with humans |
| Industrial Robot | Emergency stop as primary safety | E-stop during dynamic gait causes uncontrolled fall |
| Wheeled Robot | Graceful degradation under delay | 50ms delay causes instability during dynamic locomotion |
| Wheeled Robot | Simple odometry for localization | Leg slip, terrain variation require complex state estimation |
TABLE IV : Why Existing Robotic Security Solutions Fail for Quadrupeds
Fig. 5 : Evolution of robotic cybersecurity research (2015–2025). Events above the timeline represent vulnerability discoveries and attacks (red); events below represent defensive advances (green). The accelerating density of events from 2023 onward reflects the transition from academic research to active exploitation of deployed quadruped platforms.
| Aspect | Boston Dynamics Spot [ 1 ] | Unitree Go1/Go2/G1 [ 35 , 36 ] | Ghost Robotics Vision 60 [ 9 ] | ANYmal (ANYbotics) [ 39 ] | Xiaomi CyberDog [ 40 ] | DeepRobotics X30 [ 41 ] |
| Encryption | TLS 1.2+/HTTP2 | AES (fleet-wide) | None (critical) | Enterprise TLS | Cyclone DDS | Not disclosed |
| Authentication | JWT/ES256, X.509 | BLE (vulnerable) | None in protocol | Enterprise PKI | Default mi/123 | Not disclosed |
| Firmware Security | Signed hash trees | Limited | Not documented | Industrial cert. | Open source | Proprietary |
| Key Management | Per-device unique | Hardcoded fleet | Hardcoded in APK | Per-device | Not documented | Not disclosed |
| Physical Ports | Restricted | Open | Unauth. (critical) | Industrial sealed | Open access | Standard |
| Cloud Backend | Optional, controlled | CloudSail backdoor | Military networks | Enterprise cloud | Xiaomi servers | Proprietary |
TABLE V : Comprehensive Security Analysis of Commercial Quadruped Platforms
| Survey | Quadruped | Teleop. | TRL | Attack-Conseq. | VR/AR | Platform |
|---|---|---|---|---|---|---|
| Specific | Security | Assess. | Mapping | Attacks | Comparison | |
| Botta et al. (2023) [ 5 ] | ✗ | Partial | ✗ | ✗ | ✗ | Limited |
| Verma et al. (2025) [ 42 ] | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Yaacoub et al. (2022) [ 20 ] | ✗ | Partial | ✗ | Partial | ✗ | ✗ |
| Hamdan & Mahmoud (2021) [ 43 ] | ✗ | ✓ | ✗ | Partial | ✗ | ✗ |
| Li et al. (2025) [ 44 ] | ✗ | ✗ | ✗ | Partial | ✗ | ✗ |
TABLE X : Comparison with Existing Surveys on Robotic Cybersecurity
Fig. 6 : Six-layer attack taxonomy for quadruped robot teleoperation security. Each layer targets distinct system components with unique attack vectors. The taxonomy progresses from perception-level attacks (L1-L2) through communication and control (L3-L4) to infrastructure-level threats (L5-L6).
| Attack Layer | Attack Type | Mechanism | Target Component | Success Rate | Key References |
|---|---|---|---|---|---|
| Visual/Perception | LiDAR Spoofing | Laser injection/relay | 3D perception | 75–92% | [ 52 , 53 ] |
| Camera Manipulation | Video injection/editing | Operator view | High | [ 61 ] | |
| Visual Odometry Attack | Keypoint manipulation | Ego-motion estimation | Variable | [ 62 ] | |
| Depth Perception Attack | Stereo matching exploit | Distance estimation | 80%+ | [ 61 , 54 ] | |
| VR/AR-Specific | Cybersickness Induction | Flicker/vection manipulation | Human operator | High † | [ 66 ] |
| HMD Side-Channel | Motion/gesture analysis | Operator privacy | 82–90% | [ 72 ] |
TABLE XII : Comprehensive Attack Taxonomy for Quadruped Robot Teleoperation Communication Systems
Fig. 7 : Three primary cascading failure pathways in quadruped teleoperation. Perception-Balance cascades operate in sub-second timeframes, Operator-System cascades develop over minutes, and Communication-Control cascades can be near-instantaneous during dynamic locomotion.
| Initial Attack | Primary Effect | Secondary Effect | Terminal Consequence | Time-to-Conseq. | Intervention Points |
| LiDAR Spoofing | Phantom obstacle or removal | Emergency stop on stairs / collision | Fall + platform damage | 500ms | Multi-sensor fusion, gait-phase protection |
| Cybersickness Induction | Operator disorientation | Delayed/wrong commands | Mission abort + operator incapacitation | 2–5 min | Real-time symptom detection, FOV adaptation |
| GPS Spoofing | Wrong position estimate | Path deviation | Entry to hazardous zone / mission failure | 10–60s | IMU/GPS consistency, VO fallback |
| MITM on Video | Delayed visual feedback | Operator overcorrection | Collision with obstacle | 1–3s | End-to-end encryption, latency monitoring |
| Control Command Injection | Forced gait transition | Loss of balance | Fall + potential human injury | 100ms | Command authentication, bounded control |
| Jamming during stairs | Communication loss | Autonomous fallback fails | Uncontrolled fall | 200ms | Multi-link redundancy, safe-stance mode |
TABLE XIV : Attack Chain Examples: From Initial Attack to Terminal Consequence
| Defense Mechanism | TRL | Validation Status | Quadruped-Specific Validation | Deployment Barrier | Key Ref. |
|---|---|---|---|---|---|
| Tier 1: Field-Deployed (TRL 7–9) — Immediately Deployable | |||||
| TLS 1.2+/DTLS Encryption | 9 | Industry standard, deployed on Spot | Yes (Boston Dynamics Spot) | None; 1–3ms overhead | [ 1 ] |
| X.509/JWT Authentication | 9 | Industry standard, deployed on Spot | Yes (Boston Dynamics Spot) | Certificate management complexity | [ 1 ] |
| Firmware Signing (Hash Trees) | 8 | Deployed on Spot, standard practice | Yes (Boston Dynamics Spot) | Requires secure boot chain | [ 1 ] |
| FHSS Anti-Jamming | 8 | Mature military/WiFi technology | No quadruped-specific testing | Synchronization during dynamic gait | [ 45 ] |
| Multi-Link Redundancy | 7 | Commercial UAV/military systems | No quadruped failover time validation | Sub-200ms failover not yet proven | [ 74 ] |
TABLE XVI : Technology Readiness Level Classification of Defense Mechanisms for Quadruped Teleoperation
Fig. 8 : Defense coverage heatmap across attack layers (L1–L6) and defense mechanisms. Cell colors encode the same effectiveness levels as Table XIX : green denotes strong coverage ( ), yellow denotes partial coverage ( ), and white denotes no coverage or non-applicability (–).
Fig. 9 : Maximum defense maturity (TRL) by attack layer. Dashed lines indicate TRL 3 (proof-of-concept) and TRL 7 (field-deployable) thresholds. Layers 1, 2, and 5 lack field-deployable defenses entirely.
| Attack Class | Observable Signals | Metrics | Test Conditions | Instrumentation |
| LiDAR Spoofing | Point cloud anomalies, depth inconsistency, object hallucination | Detection rate (%), FP/hour, time-to-detection (ms) | Flat/stairs/outdoor; trot/walk gaits | LiDAR + RGB camera + IMU |
| FDIA | State estimation residuals, joint torque deviation, gait timing errors | TTD (ms), stability margin reduction (%), fall rate | Multiple gaits, varied terrain | Joint encoders + IMU + F/T sensors |
| Cybersickness | Head motion patterns, gaze instability, HRV changes | Symptom onset (min), SSQ score, task error rate | VR interface, 30+ min sessions | Eye tracker + head IMU + HRV monitor |
| Jamming/DoS | Link quality (RSSI), packet loss, latency spike | Recovery time (ms), control continuity (%), fall rate | Dynamic locomotion, stairs, multi-link | RF spectrum analyzer + robot telemetry |
| GPS Spoofing | Position jumps, IMU/GPS inconsistency, heading drift | Detection accuracy (%), max deviation (m) | Outdoor, dynamic motion | Multi-GNSS + IMU + VO |
| Command Injection | Unauthorized command detection, trajectory deviation | Detection rate (%), response latency (ms) | Teleoperation active, varied autonomy | Command logger + state monitor |
TABLE XXIII : Evaluation Blueprint for Quadruped Teleoperation Security Research
More in this topic
Cybersecurity
Cybersecurity develops methods to protect digital and cyber-physical systems from attacks, vulnerabilities, data loss, and misuse.
See all 355 papers