Fraudsters are able to successfully sell fraudulent ads by tricking fraud detection methods. Current detection tools have limitations and are easily fooled by techniques like stacking ads above the fold or using fake traffic sources. One fraud operation, Methbot, was able to avoid detection for years by disguising fraudulent publishers and simulating human behaviors. Measurement of ad views, viewability, and other metrics can also be plainly incorrect depending on tag placement. Fraud is a significant problem, especially on mobile and through fake inventory sold on ad exchanges.
Fraud Detection is Easily Fooled by Tricking Measurement and Reporting
1. Fraud Detection is
Easily Fooled
August 2017
Augustine Fou, PhD.
acfou [at] mktsci.com
212. 203 .7239
2. August 2017 / Page 1marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Bad guys’ tool to trick detection
Source: Ratko Vidakovic
3. August 2017 / Page 2marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Fraudsters successfully sell ads… how?
100% viewability
(but, it’s fake)
AD
Stack ads all
above the fold to
trick detection
0% NHT
(but, it’s fake)
Buy traffic that is
guaranteed to
pass fraud filters
clean placement
(but, it’s fake)
Pass fake source
to trick reports of
placement details
http://www.olay.co
m/skin-care-
products/OlayPro-
X?utm_source=elle
&utm_medium=dis
play
+ +
“by tricking measurement and reporting”
4. August 2017 / Page 3marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Current detection cannot catch it
In-Ad
(billions of ads)
• Limitations –
tag is in foreign
iframe, cannot look
outside itself
ad tag / pixel
(in-ad measurement)
In-Network
(trillions of bids)
On-Site
(millions of pageviews)
javascript embed
(on-site measurement)
• Limitations –
most detailed
analysis of visitors,
bots still get by
• Limitations –
relies on blacklists
or probabilistic
algorithms, least info
ad
served
bot
human
fraud site
good site
5. August 2017 / Page 4marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Just because you can’t measure it
100%
fraud
> 50% fraud
… doesn’t mean it’s not there.
6. August 2017 / Page 5marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Mobile is fraught with fraud, NOT 2%
1 2
66% avg fraud
18% avg fraud
1. 9% of the apps (blue dots) caused 52% of impressions; 66% fraudulent
2. 91% of apps caused the 48% of impressions; 18% fraudulent
• 1 billion mobile display impressions
• Nearly 1,000 apps cross referenced with SDK
7. August 2017 / Page 6marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Methbot avoided detection for years
Source: Dec 2016 WhiteOps Discloses Methbot Research
“Methbot, steals $2 billion annualized;
and it avoided detection for years.”
1. Targeted video ad inventory
$13 average CPM, 10X
higher than display ads
2. Disguised as good publishers
Pretending to be good
publishers to cover tracks
3. Simulated human actions
Actively faked clicks, mouse
movements, page scrolling
4. Obfuscated data center origins
Data center bots pretended to be
from residential IP addresses
8. August 2017 / Page 7marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Plainly incorrect measurements
Incorrect IVT
Measurement
Sources 1 and 2
measured on-page
Source 3
in foreign iframe
1x1 pixel
incorrectly reported as
100% viewable
Incorrect
Viewability
9. August 2017 / Page 8marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Tag placement yields opposite results
Tag
(in foreign iframe)
Tag
(on page)
window sizes detected
as 0x0 or 0x8 pixels correct window sizes
for ads detected
0% humans
60% bots
60% humans
3% bots
“fraud measurements could be entirely wrong, depending on
where the tag is placed and where the measurement is done.”
10. August 2017 / Page 9marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
3 bad apps eat 3/4ths of budget
flashlight app
75% of the
dark red
keyboard app alarm clock app
11. August 2017 / Page 10marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Fake inventory sold on exchanges
publisherA.com
… but, PublisherA
does NOT sell ads
on open exchanges!
“Dark Revenue” is ad revenue diverted away from
publishers, so they don’t even see it’s missing.
• Large pubs – “dark” is 1-2X ad revenue
• Medium pubs - “dark” is 5-10X ad revenue
• Small pubs - “dark” is 20-100X ad revenue
12. August 2017 / Page 11marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Facebook viewability low? No problem
“Agencies continue to accuse
Facebook. Assuming Facebook
viewability is actually as low as
20%... No problem, because it’s
still 5X more effective and sends
10-99X more humans than other
programmatic display sources!”
13. August 2017 / Page 12marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Bad media buying goes unnoticed
Most of budget wasted
between 12a – 4a
Media turned off
between 2a-5a
14. August 2017 / Page 13marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Dr. Augustine Fou – Independent Ad Fraud Researcher
2013
2014
Follow me on LinkedIn (click) and on Twitter
@acfou (click)
Further reading:
http://www.slideshare.net/augustinefou/presentations
https://www.linkedin.com/today/author/augustinefou
2016
2015