# Kashf for UBL: identity verified in seconds United Bank Limited — banking, PK. The art of smart verification: AI-powered KYC and identity checks, implemented for United Bank Limited. - **Client:** United Bank Limited (UBL), Pakistan - **Product:** Kashf, an AI-powered KYC and identity verification solution by FetchSky - **Industry:** Banking and fintech - **Documents:** Smart and Urdu CNICs; also passports, driver’s licences and other ID cards - **Technology:** AI, OCR and computer vision, with liveness detection - **Deployment:** Cloud or on-premises; mobile SDK or iframe on the client side ## In short - Kashf is an AI-powered KYC and identity verification solution built by FetchSky and implemented for United Bank Limited. - Kashf verifies a customer in three steps: it scans and authenticates the ID, checks liveness with random head-movement prompts, and matches the face on the ID to the live video. - In testing, Kashf achieved 98% accuracy on Smart CNIC cards, 92% on Urdu CNICs and 97% on face verification. ## In numbers - 3 steps: scan the ID, prove liveness, match the face - 98%: accuracy on Smart CNIC cards, in testing - 92%: accuracy on Urdu CNICs, in testing - 97%: face verification accuracy, in testing ## Overview As banking moves online, one question matters more than ever: is the person on the other side of the screen who they claim to be? Kashf answers it in seconds. It scans a government ID, checks that it is genuine, and confirms that the person holding it is real, present and a match for the photo. FetchSky built Kashf and implemented it for UBL, one of Pakistan’s largest banks, to bring secure, simple identity verification to its digital channels. ## The challenge Digital convenience brings the risk of identity theft and payment fraud, with serious consequences for customers and for banks. Fraudsters adapt as fast as the technology does, exploiting weak points in digital systems, and traditional identity checks are no longer enough to stop them. UBL needed verification that is both stronger and easier to use — a solution that could: - Meet KYC requirements and support compliance with anti-money laundering regulations. - Speed up onboarding by reducing manual document checks and paperwork. - Resist fraud, including forged documents and attempts to fool face recognition with photos or videos. - Fit the bank's own systems and policies, including its core banking platform and deployment preferences. ## How does Kashf verify someone? Three guided steps on the customer's phone: scan the ID, prove they are live, and match the face. - Three steps — **1. ID scanning and document verification:** The camera opens with an on-screen guide showing where to place the CNIC, and real-time prompts to move closer or further away until the card is in focus. Kashf recognises whether it is seeing the front or the back, captures both, and lets the customer review the captures before going on. As it reads the card, AI checks watermarks and security features for signs of tampering or counterfeiting. - Three steps — **2. Video selfie and liveness detection:** The customer looks into the front camera and follows prompts generated at random — turn your head left, right, up or down — while Kashf tracks facial movement and expression. Because nobody can know the prompts in advance, a photo or a recorded video of someone else cannot pass. - Three steps — **3. Face and expression matching:** Kashf cross-checks three things together: the face on the ID against the live video, the gestures performed against the ones it asked for, and the authenticity of the document. The customer is verified only when all three agree. - Key capabilities — **Fast and guided:** Real-time processing and clear prompts keep onboarding quick and smooth, without sacrificing accuracy. - Key capabilities — **Fraud prevention:** Real-time document validation, forgery detection and gesture recognition reduce the risk of fraud. - Key capabilities — **Integrations:** Connects with core banking, CRM systems, government databases such as NADRA, and AML and PEP screening databases. - Key capabilities — **Customizable:** Verification parameters, document types and interface elements can be adjusted to industry needs and regulatory requirements. - Key capabilities — **Flexible deployment:** Run it in the cloud for scalability, or on-premises for enhanced security. - Key capabilities — **Accuracy:** In testing, Kashf achieved 98% accuracy on Smart CNIC cards, 92% on Urdu CNICs and 97% on face verification. Results depend on dataset quality, the style of ID, image quality, lighting and the condition of the card. ## What does Kashf read from a CNIC? From the front and back of the card, Kashf extracts: - The face image - Full name - ID number - Date of birth - Document issue date - Addresses - The QR code, which carries the ID number Reading is only half the job. In the same pass Kashf checks the card’s watermarks and security features, so a forged or altered card is caught before the selfie step begins. ## Which systems does Kashf connect to? - Core banking: customer IDs are validated in real time inside account opening, so verification is part of the banking journey rather than a detour from it. - Government databases such as NADRA: cross-referencing official records makes each check more reliable and supports regulatory compliance. - AML and PEP databases: customers are screened against lists of known financial criminals and high-risk individuals. - CRM systems: verified details flow into the records the bank already keeps. ## How is Kashf deployed? On the server, Kashf runs wherever the bank needs it. On the client, it reaches customers where they already are. - Cloud: runs in the cloud, for scale and quick integration. - On-premises: runs inside the bank, where data has to stay in-house. - Mobile SDK: built into the bank’s own app. - Iframe: embedded in a web journey, such as online account opening. ## Beyond banking The same three steps serve four jobs: - Customer onboarding: quick ID checks and live video replace manual processing. - KYC and AML compliance: real-time ID verification and data validation. - Fraud prevention: document validation and gesture recognition stop identity fraud and unauthorised access. - Age verification: reliable ID checks confirm age for age-restricted services and products. Those jobs exist well beyond banks: in insurance, government and security, IT and telecom, the sharing economy, and travel and hospitality. ## How we built it FetchSky built Kashf on AI, OCR and computer vision with liveness detection, and implemented it for UBL's digital channels — deployable in the cloud or on-premises, and delivered to the customer's phone through a mobile SDK or an iframe. ## Outcome UBL now has an AI-powered way to verify customers digitally, replacing slow manual checks with a fast, guided process that is hard to fool. For customers, opening an account or accessing a service takes a few taps. For the bank, it means stronger KYC and AML compliance and lower fraud risk. ## Outcomes - Identity verification: Seconds, guided on the customer’s phone. ## Questions ### What is Kashf? FetchSky's AI-powered KYC and identity verification: it scans a government ID, checks it is genuine, and confirms the person holding it is real, present and a match. ### What does Kashf read from a CNIC? The face image, full name, ID number, date of birth, issue date, addresses and the QR code, while checking watermarks and security features for tampering. ### How does Kashf stop a photo or video from passing? Liveness detection asks for random head movements, so a static photo or a recorded video cannot pass. ### Which systems does Kashf integrate with? Core banking, CRM systems, government databases such as NADRA, and AML and PEP screening databases. ### Can Kashf run on-premises? Yes. It runs in the cloud or on-premises, with a mobile SDK or iframe on the client side. Source: https://fetchsky.com/case-studies/kashf-ubl