引導式測驗 · 中等 等級
綜合身分詐欺偵測測驗
了解合成身分、網路訊號、實體連結錯誤、延遲標籤以及合法薄文件申請人的保護措施。
問題 1 的 8
How does synthetic identity fraud differ from simple impersonation?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- How does synthetic identity fraud differ from simple impersonation?
- It uses only one real person’s complete, unaltered identity
- It always involves a stolen payment card rather than an account
- It requires the real person to approve each transaction
- It is defined by changing a model training label
- Why can synthetic identity accounts be hard to detect immediately?
- They never involve account activity
- They may build a history over time before suspicious use becomes visible
- They always use verified biometrics
- Their applications contain no information
- What can entity resolution do in a fraud system?
- Prove every linked record belongs to the same person
- Replace all identity checks
- Replace identity verification and source-data checks
- Automatically approve every account in a linked cluster
- Why should a data mismatch not automatically prove fraud?
- Legitimate life changes and data errors can create inconsistencies
- Only synthetic identities ever contain inconsistent records
- An inconsistency always means deliberate deception
- Address data cannot be compared across records
- What evaluation challenge comes from delayed fraud confirmation?
- All fraud labels are available instantly
- Rejected applications provide full repayment outcomes
- Positive labels may arrive long after the original application
- A model can use future labels at decision time
- Which metric helps evaluate applicant harm alongside fraud detection?
- Only total linked accounts
- The total number of linked records, without confirmed outcomes
- The count of flagged accounts without review results
- Average model size or hardware memory use
- Why protect identity-linkage graphs and investigation notes?
- Graphs contain no personal data
- Linkage prevents unauthorized access
- Notes are always public records
- They can expose sensitive information about many people
- How should an identity-risk model contribute to an account decision?
- Provide evidence for review while preserving correction and human-review paths
- Make unexplained adverse decisions based on one weak match
- Classify every thin-file applicant as fraudulent
- Replace all customer verification