-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlec18_ans.mhtml
1168 lines (963 loc) · 51.1 KB
/
lec18_ans.mhtml
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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
From: <Saved by Blink>
Snapshot-Content-Location: https://6858.csail.mit.edu/lec-answers/18
Subject: Answers for lecture 18 - 6.858
Date: Tue, 21 May 2018 01:42:24 -0000
MIME-Version: 1.0
Content-Type: multipart/related;
type="text/html";
boundary="----MultipartBoundary--gYWlBCgsYZVp7U1ZOw71ahKccEFtYUnjSAryyQzUic----"
------MultipartBoundary--gYWlBCgsYZVp7U1ZOw71ahKccEFtYUnjSAryyQzUic----
Content-Type: text/html
Content-ID: <[email protected]>
Content-Transfer-Encoding: quoted-printable
Content-Location: https://6858.csail.mit.edu/lec-answers/18
<!DOCTYPE html><html><head><meta http-equiv=3D"Content-Type" content=3D"tex=
t/html; charset=3DUTF-8">
<title>Answers for lecture 18 - 6.858</title>
=20
<style type=3D"text/css">
form {
display: inline; =20
}
=20
body {
color: #2E3436;
font-family: "Georgia","Liberation Serif","Droid Serif","Helvetica","Bits=
tream Vera Serif","Serif";
font-size: 12pt;
line-height: 1.5em;
margin: 0px;
}
#container {
display: block;
height: 100%;
margin-left: 100px;
margin-top: 50px;
margin-right: auto;
min-height: 100%;
overflow: inherit;
}
.dlsubmit {
}
h1 {
font-size: 30pt;
line-height: 1.5em;
text-align: left;
margin-left: -40px;
}
h2 {
color: #8BB827;
font-size: 20pt;
line-height: 1.5em;
text-align: left;
margin-left: -20px;
}
=20
a {
color: #666666;
text-decoration: none;
}
.btn {
width: 150px;
}
.textbox {
background: none;
}
pre.student-text {
white-space: pre-wrap;
}
#shell {
border-left: 5px solid #DDDDDD;
padding-left: 10px;
margin-left: 50px;
margin-top: 25px;
}
#subguide {
margin-left: 10px;
margin-top: 25px;
margin-bottom: 50px;
}
#fileform {
background: none;
position: relative;
box-shadow: none;
}
/* a fancy submit style */
input {
-moz-border-bottom-colors: none;
-moz-border-left-colors: none;
-moz-border-right-colors: none;
-moz-border-top-colors: none;
background: -moz-linear-gradient(center top , #FFFFFF, #E9E9E9) repeat sc=
roll 0 0 padding-box transparent;
border-color: #DEDEDE #BBBBBB #BFBFBF #DEDEDE;
border-image: none;
border-radius: 11px 11px 11px 11px;
border-style: solid;
border-width: 1px;
box-shadow: 0 1px 1px rgba(0, 0, 0, 0.1);
color: #464646;
cursor: pointer;
display: inline-block;
font-family: "Lucida Grande",Tahoma,Arial,sans-serif;
font-size: 100%;
line-height: 130%;
margin: 0 0.7em 0 0;
padding: 5px 10px 6px;
text-decoration: none;
}
=20
a:hover {
text-decoration: underline;
}
table {
background: none repeat scroll 0 0 #FFFFFF;
border-collapse: collapse;
margin: 10px;
text-align: left;
width: 1000px;
}
=20
th {
border-bottom: 2px solid #6678B1;
font-weight: bold;
padding: 10px 8px;
}
.spanned {
border-bottom: none;
text-align: center;
padding: 0;
}
=20
td {
padding: 9px 3px 0;
}
#header {
background-color: #EEEEEE;
height: 30px;
}
#menu {
padding-left: 10px;
font-size: 20px;
display: inline;
}
#menu a {
margin-left: 5px;
margin-right: 5px;
}
</style>
=20
</head>
<body>
<div id=3D"header">
<div id=3D"menu">
<a href=3D"http://pdos.csail.mit.edu/6.858">6.858</a>
=20
| <a href=3D"https://6858.csail.mit.edu/reset">New api-key</a>
| <a href=3D"https://6858.csail.mit.edu/logout">Logout</a>
=20
=20
</div>
=20
</div>
=20
<div id=3D"container">
=20
<h1> 6.858: Answers for lecture 18 </h1>
<ul>
<li><pre class=3D"student-text">They can prevent recursive queries on their=
domains in order to make crawling more difficult.</pre></li>
<li><pre class=3D"student-text">This study requires a lot of web crawling, =
so if the operators of the spam value chain notice lots of requests coming =
from the same source, they should probably block all requests from there, s=
ince "normal" users shouldn't make *that* many requests...=20
</pre></li>
<li><pre class=3D"student-text"> How could the operators of the spam value =
chain, studied in this paper, make it more difficult to repeat such studies=
in the future?
Taking the example of Pharmacy Express in the paper, the operators of the s=
pam value chain could take steps to hide the business chain of their spam. =
For example, they could register under a fake corporation for DNS and find =
a way to hide the bank account that the online payment would route to. This=
would mean that the people studying their spam value chain would have much=
less information about the spammers. In this paper, there is a pretty clea=
r map of where the money is flowing.
</pre></li>
<li><pre class=3D"student-text"># Lecture 18
------------------------------
How could the operators of the spam value chain, studied in this paper,
make it more difficult to repeat such studies in the future?
------------------------------
It's possible that an initial link could redirect to a CAPTCHA page, which
after completion then redirects to the ad. This would make it more costly
to crawl all the links received. As there for nearly a billion URLs crawled=
, a
CAPTCHA at each would seriously slow down crawling (although it also may hu=
rt
their sales as well).
Additionally, it seems that the botnet operators could use a TOR based botn=
et to
send emails anonymously from random TOR exit nodes. This would not hide the
click trajectories, as spam targets most likely don't use TOR, but it could
obfuscate more information about the perpetrator.
</pre></li>
<li><pre class=3D"student-text">- Spam advertisers can fan out the ASes the=
y choose to host their domain names on, and not select the ones with high d=
ensity.
- Ensure non US banks for transactions, and ensure they aren't the three ba=
nks listed in the paper. Enforcing payment blockages are weaker.
- Even in choosing registrars, ensure there's a global spread, since media=
ting/law enforcement with global resources is harder.</pre></li>
<li><pre class=3D"student-text">- Use patterns common to legitimate website=
s to hinder program tagging based
on strange hosting conventions or storefront templates
- Diversify ASes and registrar sources
- Diversify payment processing & fulfillment between different affiliat=
e programs
- Create more bots like Rustock to "blacklist poison" and make more junk da=
ta
</pre></li>
<li><pre class=3D"student-text">- hide information about the merchant bank =
=20
- introduce more variance into the HTML text on web pages to make it harder=
to cluster them. Or perhaps make some of the text into images, making it h=
arder to collect information about them.
- include captchas or other checks to prevent crawling by bots
- limit querying servers repeatedly (which the authors had to do in order t=
o get the whole set of domains used to handle clicks) </pre></li>
<li><pre class=3D"student-text">- make it harder to content cluster - chang=
e up the structure of their web pages
- more levels of URL redirects</pre></li>
<li><pre class=3D"student-text">4/24/18
Lecture 18
Question:
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
Answer:
Randomize url redirects to make it harder to web crawl extensively, use eve=
n more ip addresses for the same domain to make it harder
to track, acquire more urls, hide links so that your webpage cant be crawle=
d and only the relevant buttons (ie. to purchase) can be
clicked.</pre></li>
<li><pre class=3D"student-text">6.858 - Spam
The spam value chains can try any of these measures:
- Locate and fill honeypots/other collection sources with garbage urls
- Have servers that deny http gets to web crawlers/similar bots
- Randomize page layout to make clustering harder
</pre></li>
<li><pre class=3D"student-text">> How could the operators of the spam va=
lue chain, studied in this paper, make it more difficult to repeat such stu=
dies in the future?
They are already using a very decentralised system to avoid scrutiny,
and they fact that the value chain has been so stratified makes it harder
to analyse.=20
One aspect that they mention is that the payment infrastructure used in the
value chain is very slowly changing in comparision to the E-Mail/DNS/Hostin=
g
side of things. So finding ways to become more flexible w.r.t to banking
providers would increase resilience in the spam value chain and make it=20
harder to analyse.
</pre></li>
<li><pre class=3D"student-text">A simple way to prevent a robot from crawli=
ng is to add captchas, as this would increase time/cost of performing this =
type of analysis.
</pre></li>
<li><pre class=3D"student-text">A spam value chain operator could diversify=
the avenues through which they emit spam to make it harder to collect data=
. Email makes it really easy to collect data. Using other sources would mak=
e it harder to conduct a study. </pre></li>
<li><pre class=3D"student-text">A spamming website could impose captchas so=
that they cannot be automatically scraped by bots, which this analysis rel=
ies on. </pre></li>
<li><pre class=3D"student-text">Although significantly more difficult than =
using a botnet, a sufficiently powerful spammer could infiltrate and exploi=
t legitimate sites. This would then insert a legitimate source into the spa=
m value chain and therefore help the email from being regarded as spam in t=
he first place as the study gathered spam urls from anti-spam providers.
In addition a spam link could slow down the web-crawling portion of the stu=
dy by adding some mechanism to ensure an actual user is clicking through th=
e spam. Possibly by inserting some mechanism the user must interact with to=
gain access to the actual content (i.e. clicking a legitimate looking "ent=
er site" button) and then reject any clicks that seem non-human.</pre></li>
<li><pre class=3D"student-text">Among the different methods possible, one w=
ay to make this kind of=20
study harder in the future is for affiliate programs to make dissimilar
templates for their users. Currently, users which use the same affiliate
program will end up with very similar websites in terms of realization,
so it's easy to deduce the influence of different affiliate programs and=20
to deduce information about them.</pre></li>
<li><pre class=3D"student-text">Answer to Paper Question:
They need to obfuscate name-records more. They could use the premium hosts =
discussed in the reading to furhter obfuscate their DNS information. Furthr=
more, they could add more "blacklist-poisoning" to hamper the researher's a=
bilities. Finally, they could obfuscate URL extraction in some manner to pr=
event a bot from easily collecting this data but still letting consumers ge=
t to the website. </pre></li>
<li><pre class=3D"student-text">As time goes by, there will be deeper level=
s of spam services as the operators sign contract to let spammers do their =
work. The center of the spam will also change.</pre></li>
<li><pre class=3D"student-text">Avoid sending spam to known spam researcher=
s. </pre></li>
<li><pre class=3D"student-text">Consistently change the urls they use and m=
igrate their website to a new domain though this might be hard to do in pra=
ctice.=20
Some form of obfuscation via blacklist-poisoning
More legitimate looking templates that make it harder to decipher what a sp=
am is
</pre></li>
<li><pre class=3D"student-text">Detect physical mail addresses ordering man=
y products in a short timeframe and block these orders
</pre></li>
<li><pre class=3D"student-text">First, the operators of these websites coul=
d provide some kind of rate-limiting delay between traversing pages, slowin=
g down the web crawler.=20
The site can also attempt to identify the parser, and blacklist the associa=
ted IP address. Second, they could try and limit the ability to classify=20
these webpages by removing identifying key phrases or text, or even by usin=
g images instead which would then not be parsable. They could also attempt
to avoid displaying information that places their sites as part of a larger=
program, thus making it more difficult to target the spammers specific
platform.</pre></li>
<li><pre class=3D"student-text">Firstly, they could add a lot of randomizat=
ion to the layout of their webpages. The authors were able to associate pag=
es as belonging to certain programs because of they attributes of the under=
lying page. Perhaps these server could instead serve pages with machine-opt=
imised code and certain amounts of variation to make such association harde=
r.
Secondly, they could try to split their payment systems. The use of a small=
number of bank accounts and processors made it very easy for the authors t=
o track these programs, even when the sites appeared to be from two separat=
e affiliates. The spam operators should try to have their affiliate program=
s operate on completely different payment networks, or even incorporate oth=
er less regulated forms of payment, like cryptocurrencies. Cryptocurrencies=
like monero would also make it harder to track these payments on the autho=
rs' side.
</pre></li>
<li><pre class=3D"student-text">From my understanding, there are several wa=
ys the operators
can do:
1. Use diverse URLs or URLs that look official. Make URLs
collecting harder.
2. Use dynamic contents, which makes content clustering really
hard and impossible. Maybe have different contents for different
users.
3. Overall, the goal is to make everything hard to track and=20
not to display any certain pattern.</pre></li>
<li><pre class=3D"student-text">Given that they ignore IPv4 domains, perhap=
s hosting sites on IPv4 addresses is the way to go?
</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain studied in this paper, make it more difficult to repeat such studies i=
n the future?
The operators can have their own protections set in place that block an ip =
from cclicking their links if it has already clicked on a bunch of links (i=
e collecting 11m links that corresponded to 348 domains means that each dom=
ain had many links and their protection should prevent someone from clickin=
g on more than some threshold of links). They could also make their website=
s harder to crawl by obfusicating the website code.=20
</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it
more difficult to repeat such studies in the future?
They could rent their services to spammers on a contract basis, and let the
spammers have more flexibly of switching between different domains so that
researchers cannot track their activities.
</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
Make the chain more difficult to access by placing webstites on unreachable=
name servers.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
Blacklisting can prevent successful crawling of the spam websites, although=
it requires identification of the crawler. Operators of the spam web pages=
could try to mess up the content clustering that the study did. I'm not su=
re whether/how this is possible, but they could possibly add a lot of hidde=
n HTML text so that the clustering tool can't accurately create the cluster=
fingerprint.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
Key to their analysis is a collection of spam-advertised URLs, by setting u=
p their own honeypots or using commericial anti-spam providers. This works=
, I would imagine, because email spam is indiscriminate, and it is feasible=
to set up a honeypot that should get hit with a representative sample of a=
ll spam.
However, as operators of the affiliate marketing value chain move to platfo=
rms like Facebook, the spam (in the form of advertisements) becomes much mo=
re targeted, and it is also correspondingly more difficult to set up fake a=
ccounts as honeypots. Thus, I would imagine a broad study of the value cha=
in is much harder in the era of Facebook ad targeting.
</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
Operators could have heuristics in place to notice fast flux suppression te=
chniques (in this case the authors "[queried] servers repeatedly to enumera=
te the set of domains collectively used for click support") and remove acce=
ss to people using them. This shouldn't be actual user behavior and should=
effectively. Also, operators could just move to IPv6-only addressing whic=
h would make the namespace significantly more difficult to enumerate. This=
might make it tougher for every user to be be able to access the site thou=
gh.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
The operators can make it more difficult to repeat these studies by targeti=
ng researchers with spam and malware. Additionally, the operators could hid=
e behind more layers of obfuscation i.e. a larger network of providers that=
do not provide information
</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
They would do just the same thing anti-spammers do! Funny enough, they'd wa=
nt to verify that their users are real humans instead of botnet scrapers. E=
ven more sinister, they could try to destabilize the system that rewards pe=
rforming such studies and thereby defeat the studies at their bottleneck!</=
pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
---
Unsure of how to interfere with the web crawling process, but I would imagi=
ne that continuing to do blacklist-poisoning at increasingly greater scale =
would make the process of filtering out the 'garbage' URLs more difficult a=
nd time-consuming.
For the content clustering and tagging stage, spammers could use images for=
the frontend design instead of text that contains keywords in order to mak=
e it more difficult to classify html content automatically.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
The technique involved relied on the traceability of bank records. Using an=
onymous transfers via cyptocurrencies and buying DNS records using untracea=
ble methods as well.=20
The centralization of the current infrastructure of payment (around banks) =
and DNS (url leasing) hinders the ability of spam providers to be untracked=
. </pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
There're severar layers of protections, which the operations can apply:
1. Make it difficult/impossible to track down payments by accepting payment=
s in cryptocurrencies. Broadly speaking, it will hide the final recepients =
of the payments to.=20
2. Some possible protection from web crawling will prevent easy extraction =
of data from the websites as an attempt to recreate a holistic picture.
3. The authors mentioned that crawling all unique websites can be an unfeas=
ible exercise. Therefore, spam operators may want to create unique landing =
pages for each email they sent, though it can be an overkill taking into ac=
count the volume of the emails sent. Content delivery systems can be also o=
f a good help.
4. Tracking down the physical originators of shipping can be blocked by usi=
ng proxies - shipping handlers & fulfillment processors.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
They could use machine learning to try to detect fraudulent clicks on their=
links (ie clicks by researchers instead of by people who are actually inte=
rested customers) and direct them to alternate fake sites.</pre></li>
<li><pre class=3D"student-text">How could the operators of the spam value c=
hain, studied in this paper, make it more difficult to repeat such studies =
in the future?
One way that they could make it harder to repeat such studies in the future=
would be to routinely/consistently change their page information such as n=
aming. Further, they could provide false information about hosting or be m=
ore savvy with their URL redirection.
</pre></li>
<li><pre class=3D"student-text">I have a few ideas, but I am not sure if th=
ey would actually make repeating such studies more difficult. Perhaps using=
sites with embedded content make it more difficult to do DNS and Web crawl=
ing. Also, using images instead of text would make it more difficult to uti=
lize automatic approaches for extracting page content. Randomizing page con=
tent would also make these automatic clustering approaches more difficult.<=
/pre></li>
<li><pre class=3D"student-text">In order to make it more difficult to repea=
t a similar study of the spam monetization infrastructure, operators of the=
spam value chain could try to detect crawlers and botnets trying to sabota=
ge or snoop on the spam click trajectory chain. Spam value chain operators =
could also induce delays into purchase and/or payment processing in the rea=
lization stage -- to slow down data collection on payment tiers that could =
be useful in further studies.</pre></li>
<li><pre class=3D"student-text">In order to make it more difficult to repea=
t such studies, more spam providers
could make use of fast-flux DNS, and do things such as not use certain keyw=
ords
on their pages to make it harder for researchers to perform such automated
wide-scale studies.=20
</pre></li>
<li><pre class=3D"student-text">In section III.B of the paper, it describes=
the process of crawling.
The operators of the spam value chain could take advantage of the=20
fact that the methods in the paper don't look at every single URL,
because as described in the paper, it would be very inefficient.
</pre></li>
<li><pre class=3D"student-text">It seems that a crucial step in the data co=
llection for this paper was the DNS crawling (Figure 2 step 3). The authors=
of the spam chain could make this approach less feasible by creating a mec=
hanism on the landing site that would only accept a connection if it actual=
ly came from a human click. Captchas or other similar technology could be v=
ery effective in preventing this large-scale collection of spam website dat=
a, as it would severely reduce the effectiveness of botnets in this collect=
ion process.</pre></li>
<li><pre class=3D"student-text">It would be possible to hinder their experi=
ments if the links they used to crawl expired after some time.=20
</pre></li>
<li><pre class=3D"student-text">It's difficult to prevent studies like this=
from being made in the future. The websites hosting the front-end of spam =
(i.e. not the payment processes) are decentralized, so websites coordating =
and focusing on rules targeting a specific IP conducting the study would be=
difficult. Workarounds such as VPNs are plentiful and can also hide the id=
entity of web crawlers performing the search.=20
One possible method is to block prepaid Visa payment cards, which would dis=
courage studies from being undertaken.=20
</pre></li>
<li><pre class=3D"student-text">Jun Wan 991032935
1. The operator can use more distinct domains and try to create a better "m=
akeup" for the url link. This make the spam url link harder to detect. It a=
lso increases the workload for a researcher to DNS-crawl all the domains.
2. Spammers should use anonymous transfer network like ToR such that the re=
searcher can not tell where the spam emails come from.</pre></li>
<li><pre class=3D"student-text">Lecture 18
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
One way in which they can make it more difficult to repeat such studies is =
by not fulfilling the HTTP requests of web crawlers. If the spammer has con=
trol over the URL domain or some intermediate domain before redirect, he ca=
n deny access to crawlers.
</pre></li>
<li><pre class=3D"student-text">Lecture 18
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
They could try to blacklist any web crawling, they could maybe try to block=
their domains in some way? </pre></li>
<li><pre class=3D"student-text">Lecture 18 (Answer to Paper Question)
Question:
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
The researchers who performed the study relied on the HTML text of an adver=
tiser's website to group Web pages with similar content. To make it more di=
fficult to repeat such studies in the future, advertiser's can add hidden H=
TML elements or meaningless comments to their Web pages in order to confuse=
the clustering tool.
</pre></li>
<li><pre class=3D"student-text">Lecture 18: Security Economics
4/25/18
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
- prevent the bot from crawling through their websites by using a human ver=
ifier, sign up form, or some additional barrier
- shipping all the packages to a central distributor to prevent leaking out=
the location of the supplier
- limit the number of IP addresses the user sees by passing all the informa=
tion through a central IP address/site</pre></li>
<li><pre class=3D"student-text">Lecture 18: Spam Value Chain Reading Questi=
on
Operators of spam value chains could make it more difficult to repeat such =
studies in the future
by taking steps to significantly limit crawling efforts on their sites. The=
y could do so by incorporating
the annoying "I'm not a robot" checks that a lot of websites include now to=
validate that the party
attempting to access their site is not a machine. This could greatly limit =
crawling efforts in the future,
and could in turn prevent studies at this level from being repeated.
</pre></li>
<li><pre class=3D"student-text">Not really sure, but some way to make extra=
ction of embedded URLs more difficult; improve fast flux techniques.
</pre></li>
<li><pre class=3D"student-text">One of the weaknesses reported in the paper=
was that these spammers would often use a limited number of domain name pr=
oviders or bank institutions. The spammers could diversify making it harder=
to trace exactly where they are based or just to rely on fewer points of f=
ailure. The spammers could blacklist IP sources from the associated univers=
ities or research institutions who performed this study. The paper mentions=
trying to make purchases from some of these websites, and having them igno=
red because they thought the spammers recognized the high load of purchases=
as fraudulent -- spammers could use pattern recognition to detect similar =
purchasing patterns. Spammers already use fast flux to obscure their DNS an=
d avoid web crawlers -- more techniques like this could also make it harder=
for researchers to repeat such a study.
</pre></li>
<li><pre class=3D"student-text">One thing they could do is add in a human-v=
erifiable captcha before loading pages. This would prevent webcrawling, and=
thereby partially thwart the entire operation. Another thing that could be=
done is if too many pages are being requested by the same server (e.g. sur=
passing a threshold), refuse more requests from this server. This would als=
o prevent the webcrawling.
</pre></li>
<li><pre class=3D"student-text">One way to make studies difficult would be =
for them to blacklist honeypots in the same way legitimate parties blacklis=
t spam domains/addresses/etc.</pre></li>
<li><pre class=3D"student-text">Operators could have their botnet servers s=
hut down when taken captive.=20
</pre></li>
<li><pre class=3D"student-text">Operators of Spam Value chains could potent=
ially obfuscate certain pieces of
information from researchers like in this paper if they were aware of the
particular situation happening, but like many side channel attacks, this se=
ems
difficult to do.
</pre></li>
<li><pre class=3D"student-text">Operators of the spam value chain can look =
at the paper and examine the techniques used by the researchers and diminis=
h the usefulness
of their techniques to make it more difficult to repeat such studies in the=
future. For example, in the clustering and tagging section
of Web pages, by using the same set of words throughout each page, even for=
disparate services, while only visibly showing the pertinent
information can cause the clustering to be less effective. As another examp=
le, to minimize the usefulness of a web crawler, operators
on the spam value chain can implement a human-ness test before they show an=
y HTML content pertinent to the researchers.</pre></li>
<li><pre class=3D"student-text">Operators of the spam value chain could mak=
e it more difficult by augmenting their existing fast flux techniques and u=
sing even more IP addresses for a given domain. They could also add more in=
direction to the DNS crawling by rotating their DNS servers.</pre></li>
<li><pre class=3D"student-text">Operators of the spam value chain could mak=
e studies like this difficult to repeat in the future by making web crawlin=
g their web pages harder.</pre></li>
<li><pre class=3D"student-text">Operators of the spam value chain could mon=
itor the same feeds that were used by the researchers and immediately take =
down any urls that appear on the feed.
</pre></li>
<li><pre class=3D"student-text">Operators of the spam value chain could pre=
vent web crawling by making their domains unreachable. They could also obfu=
scate the bank codes to avoid having the banks the use be identified.</pre>=
</li>
<li><pre class=3D"student-text">Operators of the spam value chain could try=
to use more rapid domain flux to thwart the web crawler. There isn=E2=80=
=99t much they can do to stop the collection of spam-advertised URLs since =
these are meant to be widely distributed. Also there=E2=80=99s not much the=
y can do to prevent their goods from being purchased.</pre></li>
<li><pre class=3D"student-text">Perhaps spammers could also create a shared=
list of undercover buyers - something like their own "financial blacklist"=
. Once they start to notice a certain address "falling for" multiple spam a=
ttacks, they could try to block that address from accessing the research da=
ta (i.e. bank info).</pre></li>
<li><pre class=3D"student-text">Question: How could the operators of the sp=
am value chain, studied in this
paper, make it more difficult to repeat such studies in the future?
Answer:=20
Spammer can certainly use image-based websites instead of text-based websit=
es to
make automatic tagging and clustering much more difficult. Note that this
technique is mentioned in the paper -- however, not many websites currently
employ it, as mentioned in Footnote 6.
</pre></li>
<li><pre class=3D"student-text">Question: How could the operators of the sp=
am value chain, studied in this paper, make it more difficult to repeat suc=
h studies in the future?
It seems like this study was already quite hard to do given all the work to=
automate and yet they still had to manually tag/check things. One way they=
could make it harder is to separate the domains more so it makes it harder=
for researchers to cluster the same company together. Perhaps using more d=
istinc URLs could work too but it seems like they already do a good job of =
that. Also it says that they were able to cluster by similarities in hostin=
g and DOM elements so varying those would make it more difficult to repeat.=
They mention how only a few programs that use a variety of resources so if=
all sites spread out and used a little bit of every resource it would make=
it harder. Granted, it would make hosting more costly. </pre></li>
<li><pre class=3D"student-text">Quinn Magendanz
Lecture 18
How could the operators of the spam value chain, studied in this paper, mak=
e it more difficult to repeat such studies in the future?
The crawler exhaustively enumerates all A, NS, SOA, CNAME, MX, and TXT reco=
rds linked to a particular domain, so if the operators used a more diverse =
selection of records, they will go unsearched.=20
The operators could also spam for a more diverse market to make it more dif=
ficult to cluster the sites based on content.</pre></li>
<li><pre class=3D"student-text">Similar to how non-spam ISPs and email prov=
iders blacklist spam servers, spam servers could create a network that infe=
rs unreasonable use of spam sites by single IP addresses and blacklists the=
m. The url crawling was done with 100 firefox instances, likely on the same=
network - so spammers could prevent researchers from learning more about t=
heir network by filtering out addresses that crawl many associated spam sit=
es, more than any single user would.
</pre></li>
<li><pre class=3D"student-text">Since a lot of the data is obtained by craw=
ler, the operators can block access based on accessing frequency. In partic=
ular, the paper requires querying servers repeated to obtain the set of dom=
ains used for click support, which pattern the operator can observe and blo=
ck.=20
The operator can also make their website mostly purely image based, and wit=
h only very few texts --- this will confuse the content clustering process =
as it relies on text data. It might also be possible to flood documents wit=
h excess words(i.e. cat, dot etc that does not carry any negatigve categori=
zation), then make them invisible using CSS or even just hidden by javascri=
pt. So for an end user he/she won't see these words but the flood of such w=
ord will also confuse the categorizor.
</pre></li>
<li><pre class=3D"student-text">Since the crawler ignores IPv4 addresses, a=
nalysis can be stopped by using IPv4 addresses instead.</pre></li>
<li><pre class=3D"student-text">Since the operators of the spam value chain=
were disclosed in this paper, anti-spam services are likely to augment and=
evolve their own resources to account for these operators. Therefore, in t=
he future, many of these same operators will not be applicable anymore.=20
</pre></li>
<li><pre class=3D"student-text">Since the study is based mostly on data acq=
uired by means of web crawlers, it would make sense for the people involved=
with the spam value chain to make it harder for webcrawlers to function. T=
his may involve trying to filter out webcrawlers from accessing the page, o=
r maybe adding levels of complexity that would confuse the webcrawler.
</pre></li>
<li><pre class=3D"student-text">Some actions the spam operators take to obs=
truct such studies are:
* Poison the URL blacklists with random domains.
* Diversify the DOM structures to avoid automated content clustering.
* Create multiple aliases for the same brand so it=E2=80=99s difficult to i=
dentify the affiliation unless purchases are made.</pre></li>
<li><pre class=3D"student-text">Spam value chain operators could continue t=
he process of creating more and more garbage domains to make it more diffic=
ult for the studies to farm useful results. Additionally, maybe it is someh=
ow possible for websites to prevent screenshotting of their content, or oth=
erwise mask it in some way, to prevent future studies from easily automatin=
g the process of grabbing the contents of the website the way this one did.=
According to the papers, scammers also try to blacklist security companies=
who are doing the crawling, so this is another option. Storefronts could a=
lso make their page entirely images so that the text is not scannable for c=
ategorization.</pre></li>
<li><pre class=3D"student-text">Spammers could avoid being detected in such=
a study by improving their ability
to get around spam filters, which would avoid their URL getting picked up b=
y
the researcher's botfarm. They could also employ more complicated strategie=
s,
such as frequency switching domain name servers to confuse the DNS crawler.
</pre></li>
<li><pre class=3D"student-text">Spammers could modify their webpages in way=
s that are visible to crawlers but invisible to end-users; for instance, ra=
ndomize the exact patterns of capitalization/indentation/spacing/tags in th=
e HTML to confound content clustering, include irrelevant keywords in small=
fonts or non-displayed elements to confound category tagging.
</pre></li>
<li><pre class=3D"student-text">Spammers could use short DNS records to mak=
e it harder to track.
</pre></li>
<li><pre class=3D"student-text">The "attack" that the authors of this paper=
executed against the spam "system" can be outlined as:
1. Spam data collection (and analysis)
2. Crawling of naming/hosting infrastructure (and clustering/analysis)
3. "Following the money" when actually purchasing from these sites (and ana=
lysis)
To combat these "attacks," spammers can try a variety of tactics, but I wou=
ld anticipate the persistence of the researchers to quickly overcome the si=
mple changes that could beat some of the research processes.
First, the researchers already enumerate small annoyances such as redirects=
(HTML, javascript, and even Flash) and outline fairly complete nameserver =
connections with their crawler. Simple changes like new kinds of redirects =
wouldn't have much effect (the researchers are using emulated firefox brows=
ers, which are fairly standard).
Another way to make it difficult would be to add more noise to the sites, s=
o the content clustering algorithm would have more difficulty grouping adve=
rtisements.
And finally, they could mess with the selective purchasing by adding a pseu=
do-fraud protection system in order to filter out purchases from researcher=
s (though this would be extremely difficult to implement in practice).
</pre></li>
<li><pre class=3D"student-text">The authors of the spam value chain paper w=
ill make it more difficult to perform these kinds of studies in the future,=
because their paper being published will cause spammers to be more careful=
about hiding their tracks and protecting themselves from researchers who t=
ry to study their value chains and undermine their monetization. Thus, the =
attackers will defend against the methods the researchers use and make it h=
arder to conduct similar studies in the future.
</pre></li>
<li><pre class=3D"student-text">The main source of data was the countless s=
pam messages/email etc. that they were previously able to acquire.=20
The authors then would get the URLs contained in these emails and have a we=
b crawler go through them in order
to characterize the websites, IPs etc. So here's an interesting idea: since=
those links are themselves
already links to websites that will eventually transfer the user to the act=
ual website, one can have those intermediate
steps expire after a while that the spam was sent. So we would be virtually=
cutting the connection between future researchers/users=20
that would try to access the website. Another idea, although harder to impl=
ement, would be to have the operators share information across
them. So that if any username with a peak of random (from pharmacy to cyber=
crime) activity would appear, they could simply not allow the=20
realization phase to occur.
</pre></li>
<li><pre class=3D"student-text">The main way the researchers of this paper =
studied the spam chain was through web crawling. If the operators of the sp=
am chain can hide their URLs better or make it more difficult to crawl the =
web then this study would be harder to repeat. </pre></li>
<li><pre class=3D"student-text">The operators could change the domains of t=
he pages that the spam leads to. This would mean that it would be harder t=
o track the relationship or frequency of how spam is distributed.
</pre></li>
<li><pre class=3D"student-text">The operators of the spam value chain can m=
ake it more difficult to repeat such studies in the future by increasing th=
e variety of the resources they use, such as involving more banks and compa=
ny names within the chain.
</pre></li>
<li><pre class=3D"student-text">The operators of the spam value chain could=
construct a system in which the redirection process requires some amount o=
f additional computation work that would need to be performed by a machine =
before the redirect occurs. Although this additional computation would be s=
mall for a single machine with a single browser instance and thus lead to l=
ittle slowdown in the redirect process for individuals (perhaps one second =
per redirect), this would significantly slow down a machine running hundred=
s of browser instances since each redirect would require up to one second o=
f computation.</pre></li>
<li><pre class=3D"student-text">The operators of the spam value chain could=
make future studies harder to repeat by installing Captcha on their websit=
es to prevent defenders from crawling them as efficiently.</pre></li>
<li><pre class=3D"student-text">The operators of the spam value chain could=
make it difficult to repeat such studies by applying the fast flux idea to=
more than just ip addresses. A large part of the data is using the domain =
names and capturing their redirects and info. However, if the domain names =
themselves change what they point to over time just as the backing IPs do, =
it would be difficult to conduct such a study without repeatedly polling al=
l domain names in the spam content.
</pre></li>
<li><pre class=3D"student-text">The operators of the spam value chain could=
try to conceal the details of their operations, e.g. by using a hosting se=
rvice or faking a postmark on a package.
</pre></li>
<li><pre class=3D"student-text">The operators of the value chain could make=
it more difficult to repeat this study by banning IP addresses that access=
the URLs that appear to be crawlers. This could prevent ever getting to t=
he distinct domains after clicking on the Received URLs as well as accessin=
g websites that have different domain names but are run by the same entity.
</pre></li>