• Team was involved in various hackathon projects in Jan 2025, and two projects will be piloting within the ScamShield ambit
  (https://www.hack.gov.sg/2025/ai-aunty/ and https://www.hack.gov.sg/2025/unpacked/)
• Underwent penetration testing (VAPT) to identify vulnerabilities not commonly known and took proactive measures to secure our system
• Push Notifications MVP (Closing the feedback loop with users who submit scam reports)
  • We want to provide essential feedback to users while managing notification volume for frequent reporters.
  • When the authorities marks a report as scam, the server will trigger a push notification to the reporting user’s device. The notification will contain a brief message about the report status. Users can tap the notification to view a screen showing the report outcome.
• Expanded Safe caller whitelist and built trust
  • Enabled 123K checks (mobile and call blocking) to be safely classified instead of being marked as ‘unsure.’
  • Included more agencies such as MAS and expanded our safe caller whitelist.
• Growing our online presence
  • Target the demographic (39 years old & below) not readily using the ScamShield app but is overwhelmingly susceptible to falling prey to scam based on the SPF’s scam data points
  • Increase user engagement leading to higher app installs as a complementary effect while overall increasing scam protection through herd immunity
  • We launched our Instagram page in Nov ’24, which has since grown to 2K followers.

Feature work & Initiatives

• Set up new scam alert system to help our team quickly spot and warn users and stakeholders about emerging scam trends (e.g. iMessage SG Customs scam)
• Piloted a feature that shows verified caller information for government agencies like CPF, MOM and healthcare institutions
• Improved Android message filtering by showing users exactly why their messages are filtered out
• Integrated with GovTech’s SATIS team to test their link-checking system, which catches harmful links more effectively (90+% vs current 70%)
• Begun sharing scam report data with GovTech SATIS to strengthen government-wide anti-scam efforts
• Improved the ‘Check Call’ reporting process to match report call fields, helping better identify suspicious callers
• Added detection of scam phone numbers within messages to better identify potential scam content
• Started an Instagram account to share scam awareness tips, targeting young adults most vulnerable to scams

System Improvements and Security

• Implemented full testing automation for our mobile app to ensure it works smoothly for users
• Prepared our backend infra for penetration testing (VAPT)

Regular Updates and Performance Improvements

• Streamlined our app’s performance, making it run more efficiently and reducing our operating costs by 30% since August
• Updated our WhatsApp bot service to direct users to our new, improved app instead of the older version
• Improved how we handle and respond to user feedback through our operations team
• Fixed bugs in both the app and admin dashboard to ensure smoother operation

• Fixed issues and optimised key user flows following feedback from beta app release in July 2024
  • Introduced enhanced call reporting flow
  • Allow checking/reporting malformed numbers
• Released the enhanced ScamShield app to all members of public in Aug 2024 via the iOS app store and Google Playstore
• Continuous iteration on the new ScamShield Admin Dashboard with the SPF Anti Scam Command team
  • Tweaked sorting to prioritize mobile reports over other kinds of reports
  • Displayed more information to aid in their assessment such as aggregated check/report counts, reporting device ID
  • Added blacklist upload and date-range export
• Increased awareness of the enhanced app by getting media buy in to feature the new features and improvements which resulted in more than 30 print and digital media articles, interviews and op-ed pieces
• Post launch of the enhanced app, we deployed a new functionality to accept email addresses as part of SMS senderID inputs when our users alerted us to the SG-Customs scam trending on iOS and RCS chats via our feedback channel. By allowing more users to report this scam variant and collect quantifiable data which were then shared with stakeholders, we were able to help authorities act quicker in blocking such scams and also alert more members of public to the emerging scam
• Worked with the SPF Scam Public Education Office team to launch the revamped scamshield.gov.sg website, ensuring a user-centric website that would complement the WOG anti-scam efforts and better educate users in self-protection
• Worked with the SPF’s appointed media vendor for the ScamShield suite marcomm campaigns, consulting and providing designs for the enhanced app and revamped website.
• Collaborated with Singpass Anti Fraud team on data sharing pilot to assess viability of ScamShield’s crowdsourced report data playing a part in decreasing time to detection of money mule accounts and taking down these accounts before more victims lose money

• Tested the enhanced ScamShield app MVP with members of public to iterate on their feedback and improve the UIUX and value proposition of the app before developing it for productionization
• Built and released the enhanced ScamShield beta app to WOG testers and worked on fixing bugs before roll out to members of public in Q3 2024
• Researched and implemented the MVP for the ScamShield Admin Dashboard which will be used by agencies working in the scam space to mark reports sent to the enhanced ScamShield app once it is rolled out to general public
• Worked with SPF and NCPC colleagues to revamp the ScamShield suite, starting wih the with ScamShield official website
• Secured Snr Mgt approval for the enhanced ScamShield app features and functionalities

• Developed the Check and Report feature as the new anchor feature for the revamped ScamShield app and conducted user tests to validate our hypothesis.
• Rebuilt the backend to implement this new feature and enhanced features for call blocking and SMS filtering.
• Researched and implemented a ML Classifier which incorporates LLM to better accurately label large datasets and provide better outcomes for the Check feature.
• Collaborated with SPF’s Anti Scam Command to compile and label scam datasets to train the new classifier to support the app rebuild.
• Worked closely with SPF’s Scam Public Education Office and the National Crime Prevention Council to rationalise the Scam Alert website, redesign the one-stop scams portal cum website which will be launched ahead of the SPF Workplan Seminar.
• Secured DBS’ and BCS’ buy-in to work on an end-to-end production pilot using SPF and ScamShield data by Q2 2024.
• Formalised legal documents which are now boilerplate for all anti-scam projects run by GovTech and SPF in collaboration with private sector entities generally reducing the barrier for scam data sharing in general.

• Completed full system migration from v1 to v2 (across v2 of the database, APIs, classifier, Admin dashboard and app.) v2 classifier improves detection of phishing links with much greater accuracy.
• Rolled out onboarding permissions to ensure users have their permissions set up properly to ensure Call Blocking & Identification, Mobile Data Access and Background App Refresh is enabled. Since the feature was launched, there was an increase from around 23% to 80% for Call Blocking & Identification permissions and an increase from around 63% to 80% for iOS Mobile Data and Background App Refresh permissions.
• Addressed several incidents and bugs - i.e. fixed a regex error which prevented call blocking from working as it should, fixed a problem of SS App wrongly classifying OTPs and removing LIKELY-SCAM from the blacklist as it was not an accurate indicator of scam, fixed an error with Auto-report consumer (AWS ECS) service that iss responsible for consuming messages from SQS causing our auto reports to be dropped and fixed an issue where a telco’s messages were wrongly getting blacklisted.
• Enhanced system monitoring as a whole to detect the functioning of components - this would allow for detection of system components not working (which was not present previously). This includes better error logging via DataDog, adding more granular categories for report source to help with debugging/analytics/ML improvements, added deviceID to all client side HTTPS requests enabling traceability and enhancing our debugging capability for prod devices and helping us to identify if an issue is device specific or user specific, measurement of whether the call directory extension was working, detecting whether ScamShield was set as the default call screening app for Android users, increase Android permissions sampling rate to 30% for better coverage etc.
• Secured Stakeholder buy-in from SPF for a blacklist and greylist API pilots with banks and telcos to automate the provision of production order data from the police to institutions and provide early detection of scam.
• Developed relationship with WOG anti-scams teams which unlocked access to overall anti-scam policy and plans landscape and industry connections (e.g. MasterCard, telco senior leadership) for data sharing pilot.
• Secured agreement of 1 bank and completed data sharing agreement to facilitate a data sharing pilot to assess the usefulness of Scamshield’s data.

• ScamShield Bot launched on July 20th to Public

  • Total User count reached ~10K Bot Users up to date
  • Generated 11,960 completed Checking sessions and 10,638 Reports on the Bot

• Released new version of ScamShield App on iOS and Android with enhanced Classifier. Updates include:

  • Enhanced machine learning algorithm to detect phishing links in SMSes.
  • Improved blocking of scams: SMSes from unknown SMS senders are now checked against a blacklist.
  • Recognition of trusted senders: the App now prevents the blocking of SMSes from trusted senders that have been registered with local authorities.
  • Improved reporting: The App now accepts reports received from other platforms.

• ScamShield Bot released to users across Whole-of-Government and targeted user groups
  • Reached 1432 Total number of Bot users
  • Generated 1336 completed Checking sessions and 1204 Reports on the Bot after 1 month of rollout
• New SPF Dashboard generating ~4,400+ Blacklisted entities

• New SPF Dashboard was launched to the SPF-Anti Scam Command Team to review and action on reports submitted via ScamShield Bot and the ScamShield.org website

• ScamShield Android App launched to public in October 2022
  • Within 1 month, Android App was downloaded by ~100K users