For sites that use non-standard default TTL values (64, 128, 255), `--fake-autottl` doesn't work properly.

[dilluti0n]

In the case of facebook.com, the default TTL values of packets sent from the server don't appear to be one of the three standard values (64, 128, 255), which seems to cause infer_hop to malfunction.

2026-02-26 16:51:37 [DEBUG] HopTab::put: hit empty 76; HopTabEntry{ ip=57.144.188.1, state=0x01, hop=42, ts=0, meta=0x0000000000002a01 }   
2026-02-26 16:51:37 [DEBUG] HopTab::put: update 76 to HopTabEntry{ ip=57.144.188.1, state=0x01, hop=42, ts=0, meta=0x0000000000002a01 }                                                                                                                                                 
2026-02-26 16:51:37 [DEBUG] HopTab::find_hop: found 76; HopTabEntry{ ip=57.144.188.1, state=0x01, hop=42, ts=0, meta=0x0000000000002a01 }
2026-02-26 16:51:37 [DEBUG] autottl: set ttl to 41                                                                                                                                                                                                                                      
2026-02-26 16:51:37 [DEBUG] HopTab::find_hop: found 76; HopTabEntry{ ip=57.144.188.1, state=0x03, hop=42, ts=0, meta=0x0000000000002a03 }  
2026-02-26 16:51:37 [DEBUG] autottl: set ttl to 41                                                                                                                                                                                                                                      
2026-02-26 16:51:37 [DEBUG] send_split: dst=57.144.188.1 segments=[0, 1] tcp_payload_len=1380   

[dilluti0n]

Actual hop count is 13.

$ sudo traceroute -n -T -p 443 57.144.188.1 | tail -n3
11  163.77.192.234  2.279 ms 163.77.192.233  2.549 ms 163.77.192.236  2.646 ms
12  129.134.95.66  3.394 ms 129.134.92.26  6.112 ms 129.134.95.14  5.647 ms
13  57.144.188.1  2.856 ms  2.951 ms  2.702 ms

[dilluti0n]

Workaround

Use --fake-ttl with a manually chosen value instead of --fake-autottl. TTL (Time To Live) works like a distance limit; the fake packet will be dropped after traveling that many hops, so it reaches the censor but not the destination server.

In practice, a value of 5–8 is conservative enough that fake packets won't reach most servers, while still passing through the ISP-level censor. Start with:

sudo dpibreak --fake-ttl 8

If bypassing still fails, try lowering the value (e.g. --fake-ttl 5).

---

If you want to be precise, find the exact hop count with:

sudo traceroute -n -T -p 443 <server-ip> | tail -n3

And set --fake-ttl to a value smaller than the hop count to the destination.

---

Note that unlike --fake-autottl, this value is fixed and not per-destination, so fake packets may reach the destination server if the TTL is set too high. Choose a conservative value.