Exploring the Dynamics of Search Advertiser Fraud Joe DeBlasio, UC San Diego Saikat Guha, Microsoft Research, India Geoffrey M. Voelker, UC San Diego Alex C. Snoeren, UC San Diego
2
Search Ad Fraud = Deceptive Advertising Search Ad Fraud ≠ Click Fraud 3
Fraudster’s Goal: Make Money Attract Traffic Monetize Traffic Nutraceuticals Spam Phishing SEO Counterfeit Goods Website Compromise Lead Generation Search Ad Fraud Malware ˙˙˙ ˙˙˙ 4
Our Goal Provide unique view of Search Ad Fraud from inside the Bing search ad network • Current state, scale of the fraud • Bidding/advertising behavior • Impact on other advertisers 5
Dataset Ads available, shown, and clicked Recent past: 2+ years, at least 6-months old Fraud as defined/identified by Bing Algorithmic & manual reports; many manual reviews 6
What fraud is excluded? Account compromise Insignificant Borderline-deceptive Shrinks as policy evolves Successful evaders Manual reports encourage detection Bottom Line: omissions don’t change analysis 7
Ongoing Problem Millions of clicks / month O($100m)/year to Bing 10m (6 month delay allows fuller analysis) 8
Many new accounts are fraudulent... …but do not survive very long. 9
Why are fraudsters stopped so quickly? 1.0 1.0 1onIrDud 1onIrDud 0.9 0.9 Ad position FrDud 0.8 0.8 0.7 0.7 0.6 0.6 CDF CDF 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 10 -2 10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 10 -2 10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 Impressions per dDy Impressions per dDy Most are loud, but the good fraudsters are hard to separate 10
Pareto Applies 1.0 CuPulatiYe 3roportioQ of Clicks 42 Year 1 41 Year 2 0.8 0.6 0.4 0.2 0.0 10 -5 10 -4 10 -3 1% 10 -2 10% 10 -1 10 0 CuPulatiYe 3roportioQ of AdYertisers 11
12
Bidding Behavior How do fraudsters behave on the network? Do they bid differently than others? What verticals do they target? Has their behavior changed over time? 13
Brief Diversion: How search ads work Advertisers bid on keyword phrases e.g. ’red roses’ Advertisers choose how keywords match against query Exact/Phrase match: “red rose” / “red rose London” Broad matching: “buy flowers” 14
Brief Diversion: How search ads work Ads are chosen & ordered by a second-price auction where advertisers pay only when their ad is clicked Auction won by expected payout, which is based on bid and past performance 15
Bidding Behavior How do fraudsters behave on the network? Do they bid differently than others? What verticals do they target? Has their behavior changed over time? 16
How do fraudsters behave? 1.0 1.0 1.0 1.0 1.0 1.0 0.9 0.9 0.9 0.9 0.9 0.9 0.8 0.8 0.8 0.8 0.8 0.8 0.7 0.7 0.7 0.7 0.7 0.7 0.6 0.6 0.6 0.6 0.6 0.6 CDF CDF CDF CDF CDF CDF 0.5 0.5 0.5 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.4 0.4 0.3 0.3 0.3 0.3 0.3 0.3 3rolifiF FrDud 3rolifiF FrDud 0.2 0.2 0.2 0.2 0.2 0.2 FrDud FrDud FrDud FrDud 0.1 0.1 0.1 0.1 0.1 0.1 1onfrDud 1onfrDud 1onfrDud 1onfrDud 1onfrDud 1onfrDud 0.0 0.0 0.0 0.0 0.0 0.0 10 -2 10 -1 10 -2 10 -1 10 -2 10 -1 10 -1 10 -1 10 -1 10 0 10 0 10 0 10 1 10 1 10 1 10 2 10 2 10 2 10 3 10 3 10 3 10 4 10 4 10 4 10 5 10 5 10 5 10 0 10 0 10 0 10 1 10 1 10 1 10 2 10 2 10 2 10 3 10 3 10 3 10 4 10 4 10 4 10 5 10 5 10 5 1ormDlized number of Dds 1ormDlized number of Dds 1orPDlized nuPber of Dds 1orPDlized nuPber of bids 1ormDlized number of bids 1ormDlized number of bids Fraudsters reduce detection surface area by creating fewer ads/keyword bids 17
How do fraudsters bid? 'BroDd' 3roportion 'BroDd' 3roportion 'ExDFt' Bids 'ExDFt' Bids 1.6 1.6 1.0 1.0 1onfrDud 1onfrDud 1onfrDud 1onfrDud 0.9 0.9 1.4 1.4 FrDud FrDud 0.8 0.8 1.2 1.2 0.7 0.7 1.0 1.0 0.6 0.6 P'F P'F C'F C'F 0.8 0.8 0.5 0.5 0.6 0.6 0.4 0.4 0.3 0.3 0.4 0.4 0.2 0.2 0.2 0.2 0.1 0.1 0.0 0.0 0.0 0.0 10 -2 10 -2 10 -1 10 -1 10 0 10 0 10 1 10 1 10 2 10 2 0.0 0.0 0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 1.0 1.0 1orPDlized DverDge Eid 1orPDlized DverDge Eid 3roportion of Ddvertiser's bids 3roportion of Ddvertiser's bids Fraudsters prefer broad matching Fraudsters bid the default amount more than non-fraudulent advertisers. 18
Verticals Targeted Install Printer Call Our Helpline Number. Online Printer Support By Experts. Discord Free Download Latest 2017 Version. 100% Free! Instantly Download Discord Now! 75% Off C0ACH Factory Outlet Enjoy 75% Off & High Quality C0ACH Bags & Purses. Winter Sale Limited Time Offer Sale on DL Flights Ending Book DL Flight and Save upto 35% on All Flights. Sale Ending Today Lose Up To 8 Lbs 1 Week Want To Lose Weight Quickly? Burn Belly Fat Fast. No Diet, No Exercise 19
Third-Party Tech Support Install Printer Call Our Helpline Number. Online Printer Support By Experts. Support for QuickBooks 1855(US/CA) 723-9621. Support for QuickBooks by ntuitQuickBooks247 Support for Apple device Live service support for apple. Get service for all devices by servcare "Windows Update" Failed? This Solution (Works100%). (Highly Recommended). Live Help For Netfli`x Instant Help For Netfli~x Issues By Expert Technicians 20
Impersonation Chase - Online Login Access Your Online Account. Sign in to Chase Online Now. Googl Search Online Search. Search & Buy Discounted Clothes & Dresses Sling TV Channel Packages Don't Miss Your Favorite Shows. Compare Sling TV Channel Packages Now! Target - Online Shopping Store Hours & Locations. Go To Target.com Online Shopping Now. MyChart Login Page Enroll, View Test Results, Schedule Appts. Pay Bills Online 21
Impersonation? Hard to identify– likely will always be a problem 22
23
Impact on other advertisers How often do good advertisers encounter fraud? How much burden does any individual bear? What cost is incurred by competing with fraud? Not just monetary cost, but also opportunity cost. Competes with fraud = shown alongside an ad marked as fraud 24
How often do fraudsters impact others? 1.0 1.0 FrDud FrDud 0.9 0.9 1onfrDud 1onfrDud 0.8 0.8 0.7 0.7 0.6 0.6 CDF CDF 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 3roportion of iPpressions DffeFted 3roportion of spend DffeFted Fraudsters rarely compete with nonfraudulent advertisers 25
Does fraud impact cost? 1.0 1.0 1.0 1.0 0.9 0.9 0.9 0.9 Cost per click (CPC) 0.8 0.8 0.8 0.8 Average charge when ad is clicked. 0.7 0.7 0.7 0.7 0.6 0.6 0.6 0.6 CD) CD) CD) CD) 0.5 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.3 0.3 0.3 0.3 3UolifiF 1onfUDud (influenFed) 3UolifiF )UDud (influenFed) 0.2 0.2 0.2 0.2 3UolifiF 1onfUDud (oUgDniF) 3UolifiF )UDud (oUgDniF) 0.1 0.1 0.1 0.1 1onfUDud (influenFed) 1onfUDud (influenFed) )UDud (influenFed) 1onfUDud (oUgDniF) 1onfUDud (oUgDniF) 1onfUDud (oUgDniF) )UDud (oUgDniF) 0.0 0.0 0.0 0.0 10 -2 10 -2 10 -2 10 -1 10 -1 10 -1 10 0 10 0 10 0 10 1 10 1 10 1 10 2 10 2 10 2 10 -2 10 -1 10 0 10 1 10 2 AveUDge 1oUPDlized C3C (86D) AveUDge 1oUPDlized C3C (86D) AveUDge 1oUPDlized C3C (86D) AveUDge 1oUPDlized C3C (86D) Fraudulent competition modestly increases cost 26
Does fraud impact ad position? 1 2 3 4 27
Does fraud impact ad position? 1.0 1.0 1.0 0.9 0.9 0.9 0.8 0.8 0.8 0.7 0.7 0.7 0.6 0.6 0.6 CD) CD) CD) 0.5 0.5 0.5 0.4 0.4 0.4 0.3 0.3 0.3 0.2 0.2 0.2 1onfrDud (influenFed) ) with FliFks (orgDniF) 0.1 0.1 0.1 1onfrDud (orgDniF) 1onfrDud (orgDniF) ) with FliFks (influenFed) 0.0 0.0 0.0 5 5 10 10 15 15 20 20 5 10 15 20 Ad position Ad position Ad position Competing with fraud typically costs one ad position 28
Does fraud impact click-through rates? 1.0 1.0 1.0 1.0 1RnfrDud (RrgDniF) 1RnfrDud (RrgDniF) 1RnfrDud (RrgDniF) 0.9 0.9 0.9 0.9 1RnfrDud (influenFed) 1RnfrDud (influenFed) Click-through Rate (CTR) 3rRlifiF 1RnfrDud (RrgDniF) 0.8 0.8 0.8 0.8 3rRlifiF 1RnfrDud (influenFed) 0.7 0.7 0.7 0.7 N clicks / N impressions 0.6 0.6 0.6 0.6 CD) CD) CD) CD) 0.5 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.3 0.3 0.3 0.3 )rDud (RrgDniF) 0.2 0.2 0.2 0.2 )rDud (influenFed) 0.1 0.1 0.1 0.1 3rRlifiF )rDud (RrgDniF) 3rRlifiF )rDud (influenFed) 0.0 0.0 0.0 0.0 10 -4 10 -4 10 -4 10 -1 10 -1 10 -1 10 -4 10 -1 10 -3 10 -3 10 -3 10 -2 10 -2 10 -2 10 0 10 0 10 0 10 -3 10 -2 10 0 AverDge C75 fRr Advertiser AverDge C75 fRr Advertiser AverDge C75 fRr Advertiser AverDge C75 fRr Advertiser Fraudulent competition significantly reduces the odds of receiving a click 29
Takeaways Fraudsters forced to behave like normal advertisers - Bing kills loud advertisers - Fraudsters must be measured to be successful Competing with fraud has cost, but is rare - Few advertisers encounter fraud often, but means advertiser won’t receive clicks - Fraudsters’ mostly competing amongst themselves Pareto principle applies - Not just Bing– all likely have few elite fraudsters - Targeted interventions are effective 30
Thanks 31
Recommend
More recommend
