1 |
On Thu, May 21, 2020 at 10:47 AM Michał Górny <mgorny@g.o> wrote: |
2 |
|
3 |
> Hi, |
4 |
> |
5 |
> TL;DR: I'm looking for opinions on how to protect goose from spam, |
6 |
> i.e. mass fake submissions. |
7 |
> |
8 |
> |
9 |
> Problem |
10 |
> ======= |
11 |
> Goose currently lacks proper limiting of submitted data. The only |
12 |
> limiter currently in place is based on unique submitter id that is |
13 |
> randomly generated at setup time and in full control of the submitter. |
14 |
> This only protects against accidental duplicates but it can't protect |
15 |
> against deliberate action. |
16 |
> |
17 |
> An attacker could easily submit thousands (millions?) of fake entries by |
18 |
> issuing a lot of requests with different ids. Creating them is |
19 |
> as trivial as using successive numbers. The potential damage includes: |
20 |
> |
21 |
> - distorting the metrics to the point of it being useless (even though |
22 |
> some people consider it useless by design). |
23 |
> |
24 |
> - submitting lots of arbitrary data to cause DoS via growing |
25 |
> the database until no disk space is left. |
26 |
> |
27 |
> - blocking large range of valid user ids, causing collisions with |
28 |
> legitimate users more likely. |
29 |
> |
30 |
> I don't think it worthwhile to discuss the motivation for doing so: |
31 |
> whether it would be someone wishing harm to Gentoo, disagreeing with |
32 |
> the project or merely wanting to try and see if it would work. The case |
33 |
> of SKS keyservers teaches us a lesson that you can't leave holes like |
34 |
> this open a long time because someone eventually will abuse them. |
35 |
> |
36 |
> |
37 |
> Option 1: IP-based limiting |
38 |
> =========================== |
39 |
> The original idea was to set a hard limit of submissions per week based |
40 |
> on IP address of the submitter. This has (at least as far as IPv4 is |
41 |
> concerned) the advantages that: |
42 |
> |
43 |
> - submitted has limited control of his IP address (i.e. he can't just |
44 |
> submit stuff using arbitrary data) |
45 |
> |
46 |
> - IP address range is naturally limited |
47 |
> |
48 |
> - IP addresses have non-zero cost |
49 |
> |
50 |
> This method could strongly reduce the number of fake submissions one |
51 |
> attacker could devise. However, it has a few problems too: |
52 |
> |
53 |
> - a low limit would harm legitimate submitters sharing IP address |
54 |
> (i.e. behind NAT) |
55 |
> |
56 |
> - it actively favors people with access to large number of IP addresses |
57 |
> |
58 |
> - it doesn't map cleanly to IPv6 (where some people may have just one IP |
59 |
> address, and others may have whole /64 or /48 ranges) |
60 |
> |
61 |
> - it may cause problems for anonymizing network users (and we want to |
62 |
> encourage Tor usage for privacy) |
63 |
> |
64 |
> All this considered, IP address limiting can't be used the primary |
65 |
> method of preventing fake submissions. However, I suppose it could work |
66 |
> as an additional DoS prevention, limiting the number of submissions from |
67 |
> a single address over short periods of time. |
68 |
> |
69 |
> Example: if we limit to 10 requests an hour, then a single IP can be |
70 |
> used ot manufacture at most 240 submissions a day. This might be |
71 |
> sufficient to render them unusable but should keep the database |
72 |
> reasonably safe. |
73 |
> |
74 |
> |
75 |
> Option 2: proof-of-work |
76 |
> ======================= |
77 |
> An alternative of using a proof-of-work algorithm was suggested to me |
78 |
> yesterday. The idea is that every submission has to be accompanied with |
79 |
> the result of some cumbersome calculation that can't be trivially run |
80 |
> in parallel or optimized out to dedicated hardware. |
81 |
> |
82 |
> On the plus side, it would rely more on actual physical hardware than IP |
83 |
> addresses provided by ISPs. While it would be a waste of CPU time |
84 |
> and memory, doing it just once a week wouldn't be that much harm. |
85 |
> |
86 |
> On the minus side, it would penalize people with weak hardware. |
87 |
> |
88 |
> For example, 'time hashcash -m -b 28 -r test' gives: |
89 |
> |
90 |
> - 34 s (-s estimated 38 s) on Ryzen 5 3600 |
91 |
> |
92 |
> - 3 minutes (estimated 92 s) on some old 32-bit Celeron M |
93 |
> |
94 |
> At the same time, it would still permit a lot of fake submissions. For |
95 |
> example, randomx [1] claims to require 2G of memory in fast mode. This |
96 |
> would still allow me to use 7 threads. If we adjusted the algorithm to |
97 |
> take ~30 seconds, that means 7 submissions every 30 s, i.e. 20k |
98 |
> submissions a day. |
99 |
> |
100 |
> So in the end, while this is interesting, it doesn't seem like |
101 |
> a workable anti-spam measure. |
102 |
> |
103 |
> |
104 |
> Option 3: explicit CAPTCHA |
105 |
> ========================== |
106 |
> A traditional way of dealing with spam -- require every new system |
107 |
> identifier to be confirmed by solving a CAPTCHA (or a few identifiers |
108 |
> for one CAPTCHA). |
109 |
> |
110 |
> The advantage of this method is that it requires a real human work to be |
111 |
> performed, effectively limiting the ability to submit spam. |
112 |
> The disadvantage is that it is cumbersome to users, so many of them will |
113 |
> just resign from participating. |
114 |
> |
115 |
> |
116 |
> Other ideas |
117 |
> =========== |
118 |
> Do you have any other ideas on how we could resolve this? |
119 |
> |
120 |
> |
121 |
> [1] https://github.com/tevador/RandomX |
122 |
> |
123 |
> |
124 |
> -- |
125 |
> Best regards, |
126 |
> Michał Górny |
127 |
> |
128 |
|
129 |
|
130 |
|
131 |
Sadly, the problem with IP addresses is (in this case), that there are |
132 |
anonymous. One can easily start an attack with thousands of IPs (all around |
133 |
the world). |
134 |
|
135 |
One solution would be to introduce user accounts: |
136 |
- one needs to register with an email |
137 |
- you can rate limit based on the client (not the IP) |
138 |
|
139 |
For example I've 200 servers, I'd create one account, verify my email |
140 |
(maybe captcha too) and deploy a config with my token on all servers. Then |
141 |
I'd setup a cron job on every server to submit stats. A token can have some |
142 |
lifetime and you could create a new one when the old is about to expire. |
143 |
|
144 |
If you discover I'm doing false reports, you'd block all my submissions. I |
145 |
can still do fake submissions, but you'd need a per-host verification to |
146 |
avoid that. |
147 |
|
148 |
Tomas |