Training set size | Geolocation method | PGE < 10 km▲ | PGE < 20 km | PGE < 40 km |
---|
> 100 | Proposed method | 80.1% | 91.9% | 98.5% |
SLG method [21] | 60.5% | 83.6% | 94.1% |
TNN method [23] | 37.2% | 83.7% | 94.9% |
> 300 | Proposed method | 84.3% | 95.2% | 99.5% |
SLG method [21] | 55.4% | 79.7% | 90.1% |
TNN method [23] | 35.7% | 81.4% | 94.2% |
> 500 | Proposed method | 86.7% | 97.2% | 99.9% |
SLG method [21] | 58.0% | 81.7% | 91.2% |
TNN method [23] | 34.4% | 78.5% | 94.7% |
- ▲“PGE < X” is short for “proportion of the entities within geolocation error being X”.
- The experimental results show that the proposed method can achieve street-level geolocation for the given IP of covert communication entity. Compared with the existing typical geolocation methods, the proposed method improves the deficiency that similar delays do not necessarily mean close geographical locations of the IPs, thereby improving geolocation accuracy. This is because the existing typical methods only rely on the delay for geolocation, while the delay in the network only has the significance of distance. The similar delays do not necessarily mean close geographical locations of the IPs, because their measurement paths may be completely different, which makes the existing typical methods less reliable. The proposed method combines the distance meaning of delay and the direction meaning of path to estimate the location of the entity. The entities with similar delay and path must have similar geographical location, so as to solve the above problem and improve the reliability of location