ID Verification and Background project.pptx

Identity Verification and Background Checks

1. 1 DELTA CAPITA PRESENTED BY AMAREKOO IDENTITY VERIFICATION AND BACKGROUND CHECKS

2. Sequence Introduction Aim Demographics Image Quality, Integrity, Face Detection and Visual Authenticity Analysis of Data Integrity Suggestions

3. Introduction 3 KY Co is a leading provider of digital Client Lifecycle Management (CLM) solutions Provides Know Your Client (KYC) reviews as well as client onboarding and refresh operations as a managed services . KY Co secured a multi-year managed services agreement with a new and upcoming UK challenger bank.

4. Introduction 4 New prospective customers of the UK challenger bank have to go through a KYC process They submit a government-issued photo ID and a facial picture KY Co performs Document check and Facial Similarity check Customer will “pass” the KYC process if the results of Document and Facial Similarity checks are “clear ” During the COVID-19, number of prospective customers increased whilst the pass rate decreased

5. Aim This work investigates the root cause why the pass rate of clients has dropped with substantial increase in number of clients. 5

6. Distribution of Clear & Rejected Cases 6 Rejected Cleared

7. Analysis of Applicant User ID 7 Out of total 5880 clients, 32 applicants were processed twice Effective data entries to be 5848 unique applicants. Of these 32 applicants, 09 applicants were cleared in both 1st and 2nd attempt 1 4 applicants were rejected in both attempts. 09 applicants were those who were rejected in 1st attempt but were cleared subsequently in 2nd attempt .

8. Clients’ Demographics 8

9. Clients’ Gender vis-à-vis Results 9

10. Clients’ Document ID Type 10 Driving License : 25.8% National Identity Card : 30.1 % Passport : 24.34% Residence Permit : 1.64% However , 377 (6.4%) applicants who were cleared did not present any documents as said field was blank while 513 (8.7%) applicants who were rejected also did not present any document ID.

11. 11 Analysis of Image Quality, Image Integrity, Face Detection and Visual Authenticity

12. Image Integrity 12 Image Integrity variable asserts whether the document was of sufficient quality to verify. 137 of applicants who were rejected in final processing were cleared 1337 applicants out of total 1474 rejected. Results Image Integrity Clear Consider Total Clear 4406 4406 Consider 137 1337 1474 Total 4543 1337 5880

13. Face Detection 13 Face detection asserts whether the face in the document matches the face in the live photo or live video contained within them were sufficient to perform a face comparison. During face detection process, almost 100% of the cleared results passed the face detection except 1, which was a blank entry. Face detection cleared 562 applicants of which 443 were later rejected during image integrity check. Data of 889 applicants (15.1% of total applicants), who were rejected in final result, was blank .

14. 14 Image Quality asserts to identify images that use sophisticated counterfeiting techniques, or the image is of poor quality (blurred, low resolution, obscured, cropped, or held at an unreadable angle ). There were 834 cases where an image was unidentifiable no matched were found. Of 640 cases cleared in image quality but rejected in final result, 137 were also cleared in image integrity while remaining 503 were rejected in image integrity. Image Quality

15. 15 448 cases were cleared which were rejected during KYC process. All of these 448 cases were also cleared in face detection and image quality However , image integrity cleared only 51 of these cases while remaining 437 were rejected during image integrity check. Of 889 cases with blank entries in visual authenticity, all these cases were also blank in face detection check, while 834 of these cases were unidentified in image quality whereas remaining 54 cases were cleared by image quality. Visual Authenticity

16. 16 Summary of Rejection Contributions

17. 17 Analysis of Data Integrity Checks

18. 18 Analysis of Rejections in Sub-Results

19. 19 Of these 889 rejections in sub-results, no record was found against variables Conclusive Document Quality, Colour Picture, Data Validation, Data Consistency, Data Comparison, Police Record and Compromised Document checks. However , of these 889 rejections, 823 were cleared in Support Document check, 55 were unidentified and 11 were blanked. Analysis of Rejections in Sub-Results

20. 20 In suspected cases, major issue occurred due to non-availability of record in data comparison. Data Comparison Result asserts whether data on the document is consistent with data provided by an applicant. Similarly , data for more than 50% of suspected cases in relation to compromised document was not available. Analysis of Suspected Rejections in Sub-Results

21. 21 Analysis of 525 cautioned cases revealed that Data Comparison emerged out to be a weak area as record of 520 cases out of 525 cases was blanked. This was followed by 440 rejections in Conclusive Document Quality. Since Conclusive Document Quality is closely related to Image Integrity as it asserts if the document was of enough quality to be able to perform a fraud inspection. Therefore , all the cases rejected in Conclusive Document Quality were also rejected in Image Integrity. Analysis of Cautioned Rejections in Sub-Results

22. 22 Suggestions

23. 23 Suggestions Image Integrity appeared to be the major area of concern as 1337 rejections out of total 1474 applicants were due to Image Integrity. This could be one of the reasons for failed Image Integrity. Therefore , all the rejections in Conclusive Document Quality were also reflected in Image Integrity rejections Develop specifications with regard to image quality for government issued photo ID and facial picture to the clients so that inadvertent and unintentional tampering or poor quality does not result in applicant rejections. 1

24. 24 Suggestions Another weak area identified during data integrity checks was Data Comparison This was the major reason detected in both suspected and cautioned cases. Data is extracted from the document through OCR for Data Comparison. OCR equipment may not be able to read information from the documents Therefore , if required OCR equipment could be upgraded 2

25. 25 Suggestions Clients may be advised to provide image as per required resolution and specifications. Second , fields in the data form submitted by the applicants may not be self-explanatory Working staff at the helpline and front desk may be trained to help clients in filling out required / mandatory fields. Additionally , forms may be revised to make them easily understandable and self-explanatory to avoid unintentional mistakes by clients while filling up data. 2

26. 26 Suggestions Although more than 90% of the suspected and cautioned cases cleared Police Record check, there were 377 and 513 entries for Document ID type were missing for cleared and rejected cases respectively. Additionally , blank entries were observed in Visual Authenticity, Face Detection, Data Comparison, Data Consistency and Compromised Document This speaks of poor manual data entry practices. O perators may be trained and educated on importance of correct and complete data entry so that right decisions may be made for clearing and rejecting applications. 3

27. Thank you