Table 2 Comparative analysis of indoor localization studies (2020–2025). The accuracy values are presented exactly as reported in the original studies. As different works adopt diverse metrics–such as horizontal or vertical error (in meters), relative improvements, percentages, or qualitative descriptions– no post-standardization was applied in order to preserve the fidelity of the original results. Readers should interpret the values in the context of each study’s methodology and evaluation criteria.
From: Comprehensive analysis of security threats and privacy issues in indoor localization systems
Study | Attack type | Defense method | Dataset | Accuracy | Privacy | Performance |
|---|---|---|---|---|---|---|
Ciftler et al.59 | Privacy breach | Federated learning | Real | 1.8 m | ✓ Strong | Low (scalability issues) |
Ko et al.58 | MAC spoofing | Random forest filtering | Real | Improved vs baseline | ✗ | Medium |
Li et al.55 | Malicious check-ins | Fingerprinting + AP subset | Simulated + Real | High | ✗ | Medium |
Li et al.54 | Fraudulent check-ins | Optimal boundary + LSM | Simulated | High | ✗ | Medium |
Nieminen et al.64 | Privacy breach | Secure two-party computation | Real | 2.2 s query time | ✓ Moderate | High |
Shubina et al.61 | Privacy vs Accuracy trade-off | Obfuscation control | Real | Moderate | ✓ Moderate | Medium |
Yan et al.62 | Physical-layer spoofing | RSSI-based detection | Real | 99.8% | ✗ | Medium |
Zhang et al.60 | Privacy exposure | Paillier encryption | Simulated | Efficient (no exact error) | ✓ Moderate | High (processing cost) |
Ambalkar et al.69 | Adversarial ML | PGD + MIM + Defense | Simulated | Good (exact N/A) | ✗ | High |
Beko et al.78 | Spoofing | WCM + GTRS bisection | Simulated | Improved | ✗ | Low |
Dervicsouglu et al.72 | Security comparison | UWB vs BLE | Real | UWB: 0.43 m, BLE: 1.54 m | ✗ | Medium |
Min et al.77 | Privacy leak | 3D geo-indistinguishability | Simulated | Good (no error given) | ✓ Strong | Medium |
Na et al.71 | Cross-tech impersonation | Detection by power variance | Real | >20 m error | ✗ | Low |
Njima et al.75 | Data scarcity | GAN + Semi-supervised | Sim + Real | 21.7%/15.3% \(\uparrow\) | ✗ | Medium |
Patil et al.68 | Adversarial ML | Adversarial training + DNN | Simulated | 84.18% | ✗ | High |
Wang et al.70 | General security | Multi-task learning | Real | <2 m | ✗ | Medium |
Boora et al.82 | Adversarial ML | Neural ODE + Adversarial defense | Simulated | High | ✗ | High |
Fathalizadeh et al.81 | Anonymization | k-Anonymity + Dijkstra | Sim + Real | Moderate | ✓ Moderate | High |
Gao et al.87 | Data privacy | FL (FedLoc3D) | Real | Improved | ✓ Strong | Medium |
Han et al.86 | Spoofing/Faulty Sensors | CNN/ResNet filter | Real | High | ✗ | Medium |
Wang et al.85 | First-order adversarial | AdvLoc (DCNN) | Simulated | <1 m | ✓ Moderate | Medium |
Yang et al.83 | Adversarial ML | SecureSense | Simulated | High (not exact) | ✗ | High |
Ye et al.84 | Adversarial APs | SE-loc semi-supervised | Simulated | 8.9 m | ✓ Weak | Medium |
Zhang et al.79 | Privacy leakage | FL + DP (Adp-FSELM) | Real | 2.22% MAE | ✓ Strong | Low |
Casanova et al.91 | Tracking | Zero-knowledge ABA | Real | Secure Auth (no loc error) | ✓ Strong | Medium |
Spoofing | UnSpoof (UWB + ToA) | Real | 30 cm | ✓ Strong | Medium | |
Kalpana et al.93 | Node attacks | 3D DV-Hop + Cryptography | Simulated | <2m | ✓ Strong | High |
Mitchell et al.90 | Adversarial ML | Adversarial training + Outlier detection | Simulated | Improved vs baseline | ✗ | High |
Mohsen et al.92 | Privacy leakage | PassiFi (DL + TDoA) | Real | Sub-meter | ✓ Strong | Medium |
Peterseil et al.88 | Signal tampering | Autoencoder + Trust score | Real | 50% RMSE reduction | ✗ | Medium |
Shakerian et al.89 | Privacy, Tampering | Blockchain + IMU + ZUPT | Real | 1.2 m | ✓ Strong | High |
Xiao et al.96 | OTA adversarial | FooLoc perturbations | Real | 70–90% attack success | ✗ | High |
Eshun et al.114 | Data leakage | Cloud Offload + Crypto | Real | Good | ✓ Strong | Medium |
Etiabi et al.112 | Communication privacy | Federated distillation | Simulated | Good (no value) | ✓ Moderate | Low |
Fathalizadeh et al.5 | Privacy | Survey + Framework | N/A | N/A | ✓ Strong | N/A |
Gufran et al.113 | Adversarial ML | CALLOC + Curriculum FL | Simulated | 6\(\times\) Error Reduction | ✓ Strong | Medium |
Hemkumar et al.118 | Geo-inference | LDP + Clustering | Real | Good (empirical) | ✓ Strong | Medium |
Li et al.119 | Tracking | RFTrack + RL agent | Simulated | Improved | ✗ | Medium |
Machaj et al.117 | AP spoofing | KNN accuracy degradation | Real | Impacted | ✗ | Low |
Moghtada et al.116 | Privacy leakage | DPGAN | Simulated | Balanced | ✓ Strong | Medium |
Boudlal et al.130 | Passive tracking | Wi-Fi CSI + DL | Real | 26.4 cm | ✓ Moderate | Medium |
David et al.128 | BLE beacon privacy | Randomized ID timing | Real | Tracked avoidance | ✓ Moderate | Low |
Nie et al.131 | User identification | MAC de-randomization + DR.LIE | Real | 1.15 m | ✗ | Medium |
Abuhoureyah et al.127 | Signal distortion | CSI-enhanced HAR analysis | Literature review | Not specified | ✗ | Medium |
Li et al.129 | Location query privacy | TEE + RNN + Key revocation | Real | <1 m | ✓ Strong | Medium |
