summaryrefslogtreecommitdiffhomepage
path: root/dpibreak.1.in
blob: 4d99dfa39efce3f94447041defd7105b25580397 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
.TH DPIBREAK 1 "March 2026" "DPIBreak {{VERSION}}" "User Commands"
.nh
.ad l

.SH NAME
dpibreak \- circumventing deep packet inspection

.SH SYNOPSIS
.B dpibreak
.RI [ OPTIONS ]

.SH DESCRIPTION
\fBDPIBreak\fR is a tool for circumventing Deep Packet Inspection
(DPI) on HTTPS connections. It is only applied locally and no other
external connection is needed.

Mainly it fragments the TCP packet carrying the TLS ClientHello so
that certain DPI devices cannot extract the Server Name Indication
(SNI) field and identify the destination site. Other method can also
be used along with it, and it is described in the \fBOPTIONS\fR
section.

DPIBreak registers firewall rules to handle inbound and outbound
packets. The rules are automatically added on startup and removed on
exit, making it effective system-wide without manual
intervention. Firewall rule cleanup relies on
\fBSIGTERM\fR/\fBSIGINT\fR. If the process is killed with
\fBSIGKILL\fR, cleanup will not occur. However, the registered nfqueue
rules simply pass packets through when no process is consuming the
queue, so this is not a concern in practice. In that case, restarting
and gracefully stopping DPIBreak will clean up the leftover rules.

To register firewall rules and verdict packets, root privilege is
required on Linux (\fBnft\fR(8) or
\fBiptables\fR(8)/\fBip6tables\fR(8) with \fBxt_u32\fR must be
available); administrator privilege is required on Windows
(WinDivert64.sys and WinDivert.dll must be in the same directory as
dpibreak.exe).

This only applies to TLS-based connections (HTTPS). UDP/QUIC (\fBRFC
9000\fR) is not affected.

.SH OPTIONS
.TP
.BR \-d ", " \-\-daemon
Run as a background daemon. Logs are written to
.B /var/log/dpibreak.log.
If a daemon is already running, it will fail with "unable to lock pid
file". To stop it, run
.B kill \(gacat /run/dpibreak.pid\(ga
as root.
.RS
.PP
On Windows, this option enters the service controller entry
point. Example:
.B sc create dpibreak binPath= \(dqdpibreak.exe \-d\(dq start= auto; sc start dpibreak
.RE
.TP
.B \-\-delay\-ms \fI<u64>\fR
Delay in milliseconds to apply between fragmented pieces of the
ClientHello. Typical values are 0–1000; larger values may increase
handshake latency. (Default: {{DEFAULT_DELAY_MS}})
.TP
.B \-\-fake
Enable
.B fake
ClientHello packet injection before sending each packet
fragmented. TCP/IP header fields follow the original packet unless you
override it using the
.B \-\-fake\-*
options described below. Packets are transmitted in an interleaved
order: (fake 1), (orig 1), (fake 2), (orig 2), ...
.TP
.B \-\-fake\-ttl \fI<u8>\fR
Override ttl (IPv4) / hop_limit (IPv6) of
.B fake
packet. Implicitly enables
.BR \-\-fake .
(Default: {{DEFAULT_FAKE_TTL}})
.TP
.B \-\-fake\-autottl
Automatically infer the hop count (TTL/Hop Limit) to the destination by
analyzing the SYN/ACK packet. It assumes the server's initial TTL is
either 64, 128, or 255. The inferred value is used for
.B fake
packet transmission to ensure they reach the censor but not the
destination. By default, it uses a delta of 1. If the hop count cannot
be determined, it falls back to the value specified by
.BR \-\-fake\-ttl .
Implicitly enables
.BR \-\-fake .
.TP
.B \-\-fake\-badsum
Corrupts the TCP checksum of
.B fake
packets. When enabled,
.B fake
packets cannot pass through most routers and will not behave as
expected. It can be useful if your router/firewall provides an option
to disable TCP checksum verification. Implicitly enables
.BR \-\-fake .
.TP
.BR \-o ", " \-\-segment-order " \fI<u32,u32,...>\fR"
Specify the order in which TCP segments of the TLS ClientHello are
transmitted. The argument is a comma-separated list of byte offsets,
which must include
.BR 0 .
Offsets are sorted to define segment
boundaries: for example,
.B 0,1,3,5
defines segments
.BR [0,1) ", " [1,3) ", " [3,5) ", and " [5,end) .
The transmission order follows the order given on the command
line. Segments whose start offset exceeds the payload length are
silently skipped. (Default: {{DEFAULT_SEGMENT_ORDER}})
.TP
.B \-\-queue\-num \fI<u16>\fR
.Linux only.
NFQUEUE number to attach to. The same queue number is
used for IPv4 and IPv6. (Default: {{DEFAULT_QUEUE_NUM}})
.TP
.B \-\-nft\-command \fI<string>\fR
.Linux only.
Custom nftables command to be executed. (Default: {{DEFAULT_NFT_COMMAND}})
.TP
.B \-\-log\-level \fI<debug|info|warning|error>\fR
Set the logging level (Default: {{DEFAULT_LOG_LEVEL}}).
Aliases:
.BR warn " \-> " warning ,
.BR err " \-> " error .
.TP

.B \-\-no\-splash
Disable splash messages at startup.
.TP

.BR \-h ", " \-\-help
Show usage information and exit.

.SH EXIT STATUS
.TP
.B 0
Successfully terminated.
.TP
.B non-zero
The program encountered an error during initialization or runtime,
resulting in an abnormal exit.

.SH EXAMPLES
.PP
Run with default options:
.PP
.RS
.B dpibreak
.RE
.PP
Run as daemon with fake ClientHello injection:
.PP
.RS
.B dpibreak \-e \-\-fake\-autottl
.RE
.PP
Send the TLS ClientHello with the second byte first, followed by the
remainder, then the first byte:
.PP
.RS
.B dpibreak \-\-segment\-order 1,2,0
.RE
.PP
This causes DPI equipment to receive an incomplete record header in
the first observed segment, preventing SNI extraction, while the
destination server reassembles the segments normally via TCP sequence
numbers.
.PP
Send segments in a custom out-of-order sequence with a delay:
.PP
.RS
.B dpibreak \-\-segment\-order 5,1,3,0 \-\-delay\-ms 10
.PP
.RE
Run with a 10 ms delay between fragmented packets and verbose logging:
.PP
.RS
.B dpibreak \-\-delay\-ms 10 \-\-log\-level debug
.RE
.PP
Use a custom NFQUEUE on Linux:
.PP
.RS
.B dpibreak \-\-queue\-num 3
.RE
.PP

.SH FILES
.TP
.I /run/dpibreak.pid
Whether running as a daemon or not, DPIBreak will fail through this
file if another instance is already running on the system.
.TP
.I /var/log/dpibreak.log
Only for daemon. log goes here.

.SH BUGS
There are two types of bugs:
.RS 4
a. DPIBreak does not work as described in the manual.
.br
b. It works as described but fails to bypass the DPI.
.RE
Reporting both cases to the bug tracker helps improve the program.
For case (b), it would be helpful if you could include information
such as your region and ISP.

Any other minor improvements or suggestions are also welcome.
.PP
You can view the known bugs list and submit reports at:
.br
.I https://github.com/dilluti0n/dpibreak/issues

.SH SEE ALSO
.BR nft (8),
.BR iptables (8),
.BR ip6tables (8)

.SH AUTHOR
Written by Dilluti0n <[email protected]>.