Ai Editorial: 5 reasons why data is set to play an integral part in combatting fraud

First Published on 14th December, 2018

Ai Editorial: Data is going to be a key weapon in the arsenal of airlines as the industry attempts to fight emerging fraud threats, writes Ai’s Ritesh Gupta.

 

Airlines acknowledge that they need to be in a position to probe as many data source as possible in order to improve the probability of uncovering and combatting fraudulent activities and transactions.

Going forward, airlines not only need to focus on their own unique data, but they also have to count on external data plus be open to collaborating with other stakeholders to stop fraudsters’ malicious moves.  

1.     Tracking and consolidating own data: Blending all the available transactional data into a single system and analysis model is critical considering where the industry stands today with CNP purchases and e-commerce sales. In addition to ticket-related revenue generation, keeping a vigil on frequent flyer miles, loyalty points, gift cards etc. is must. Considering the way fraud evolves, airlines can’t ignore options like e-gift cards. Fraudsters are capable of breaking through gift card codes through various methods such as phishing or social engineering. Airlines’ own data, especially on their own channels like a website, is important to refine analytics around it.

Big data is first used to collect information about the user’s behaviour on the website (for instance, how the mouse moves, words per minute etc.), and this information is combined with machine learning, which uses pattern recognition to map the pattern of his behaviour to match it either with positive (genuine) or negative (fraudulent) behaviour, as well as predictive analytics that records the positive/ negative behaviour and uses that on future transactions for potential signs of fraud. After the point of data collection, airlines have to amplify and triangulate the data, analysing the data through multiple permutations and combinations so as to better understand the fraud patterns left behind by fraudsters in their attempt to brute force the system.

Real-time data from airline.com can also help in curbing fraud. Blacklists rarely work because hackers will never use the same credit card information twice, while white-lists are inaccurate since white-listed customers can be compromised anytime. Real-time machine learning can help against blanket blacklists and white-lists by focusing on the customer’s behaviour instead. It works with real-time live data collected on the merchant’s website, where the system trains itself with each incoming transactions to identify fraud patterns instead.

2.     Blending data from other sectors for the benefit of airlines: Specialists serving the travel industry state that the fraud-related issues must be confronted collectively. There is strength in numbers and insight in data—and help is available to leverage them both. Specialists like Accertify are working on airline-specific offerings, and their machine learning technology aggregates and transforms information from a diverse set of sources to identify emerging fraud risks and attacks. External data can complement and lend a new dimension to internal data sources, offering a better view of shoppers and the authenticity of transactions. Evaluating IP addresses, credit card data, and email addresses can enhance a carrier’s interpretation of who is doing what—and from where they are doing it.

3.     Accuracy of machine learning: The collection of more and relevant data would help to improve the accuracy of the machine learning models by churning the data through various permutations and combinations to identify potential fraud patterns. However, ultimately a multi-disciplinary approach, that combines machine learning and other techniques to make sense of the score automatically, is required to fully automate the fraud screening process. Machine learning models are only able to provide a fraud score, of which a bulk of transactions are automated but humans are still required to review a good number of transactions that are considered borderline. 

4.     Authorization rates: Among the other areas, data is being relied upon for improving upon the authorization rates. As highlighted by Adyen, on average, 5%-15% of ecommerce credit card transactions are rejected by issuing banks, and out of these, a quarter don’t work due to shortage of convincing reasons, mostly due to old and inefficient systems. And in certain markets, authorization rates across issuers take a dip because of suspicion of fraud. In this context, it is imperative to bank on data to evaluate the main reasons behind those declines and take appropriate initiatives. For instance, one areas that could be looked upon is - issuer-specific authorization rate trends. These actions may include optimizing the type of data submitted or identifying optimal routing for a given transaction.

5.     Collaboration: A shared database or working together with relevant partners is going to be the biggest factor in combating fraud. IATA Perseuss allows members to check suspect transactions against a community database holding records from around the world. Still there is plenty to learn from other industries or law enforcement in a particular market that has managed to control fraud to an extent. With a partnership featuring different players from the industry, the government and law enforcement agencies, fraudsters are being punished. For instance, the Banking Protocol scheme in the U. K. allows bank branch staff to immediately alert police and Trading Standards if they suspect fraudulent activity. The Dedicated Card and Payment Crime Unit (DCPCU), backed by the finance industry, made 84 arrests and interviews under caution in the first half of 2018, which led to 26 fraudsters being convicted. As for capitalizing on data, intelligence is also shared with law enforcement including the National Crime Agency. A campaign is being led by Financial Fraud Action UK to help everyone protect themselves from preventable financial fraud and is being delivered with and through a range of partners in the UK payments industry, financial services firms, law enforcement agencies, telecommunication providers, commercial, public and third sector organizations.

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Why even one data breach today is enough to shake a consumer?

First Published on 4th December, 2018

Imagine receiving an email early in the morning, stating that your personal data has “possibly” been compromised. It’s disturbing. But the agony doesn’t end here. Ai’s Ritesh Gupta explains why.  

 

A data breach is a big concern for all. From consumers’ perspective, if one understand the tricky situation it can shake one and all. Come to think of it - if a user has created a formidable password and believes everything is fine, how about unearthing the fact that if there is a data breach, then the same password is leaked and is of no use. Considering what all is at stake, the leaking of passwords and how a fraudster can benefit from it is an annoying as well as distressing situation for the user.  

As experienced from an email from Quora today, the company shared that some user data was “compromised by a third party who gained unauthorized access” to its systems. Information that may or may not have been compromised includes account and user information (e.g. name, email, IP, user ID, encrypted password, user account settings, personalization data). Even though the company acknowledged that it is their responsibility to make sure things like this don’t happen, it also stated: “while the passwords were encrypted (hashed with a salt that varies for each user), it is generally a best practice not to reuse the same password across multiple services, and we recommend that people change their passwords if they are doing so”.

 

A massive problem for consumers

The agony prolongs after this email. The thought of using one password for multiple accounts haunts. 

A frail approach towards password management is enabling hackers to gain access to confidential information.

As per an analysis, initiated by password management specialist LogMeIn’s LastPass, nothing much has changed over the last two years when it comes to choosing and handling of passwords.

As we highlighted in one of our reports, consumers stick to same passwords and don’t change them often. This is a significant revelation as password stealing means all account-based online services are under a threat.

In an interview, an executive from Sift Science even pointed out that every one’s credentials have already been compromised and the industry has actually reached the point of no return. It might not be a straightforward task to gain access to everyone’s account, but just like solving a puzzle or putting several pieces together, fraudsters can sneak through the defence. So from one data beach one can get a vital piece of information about users. And then another breach sharing more details about users and so eventually cracking all details of one account.

This issue of same password for multiple accounts is a tough habit to break. Even the millennials, a group supposedly well-versed with technology, mostly reuse passwords because of fear of forgetting and commonly use a variation of 1-2 passwords they can remember! On the positive side, more users are opting for more secure password storage and automated password resets to overcome the anxiety of failing to recall, but it is a long way to go.

How to go about it?

Even as credentials are being stolen, it is imperative for organizations to bolster the authentication process. Merchants should aim to mitigate the damage done by ensuring that the stolen data cannot be used. One way to achieve this is to deploy real-time active surveillance on every login to filter out potential threats and prevent attackers from gaining unauthorized access to accounts. Organizations can avail offerings that can spot passwords that are currently in use in a domain but have been exposed in a previous data breach. As much as merchants need to take action and ensure that data doesn’t get stolen (How to prevent “Starwood guest database breach” -like incidents?), consumers, too, need to be informed.

So inform and educate customers about the significance of passwords. There might not be anything new in these instructions but nevertheless the importance of strong passwords and changing them from time to time can help. For instance, working out unique passwords that include a sequence of upper and lowercase letters, numbers and special characters. Directing users not re-use same passwords. Train the user to be security-minded and to spot scams. Also, as in case of certain apps, the password expires after a while and users are left with no option but to change it. No one likes friction in any user session, but at the end of the day the problem is too big to ignore.

 

Follow Ai on Twitter: @Ai_Connects_Us

 

 

 

 

 

Ai Editorial: How to prevent “Starwood guest database breach” -like incidents?

First Published on 3rd December, 2018

Ai Editorial: One question that organizations need to dig deep into is – how to go for end-to-end protection for the sensitive data an organization has and how to prevent a data breach? Ai’s Ritesh Gupta looks into it.  

 

It’s frightening. The number as well as the scale of data breaches is now large enough to scare possibly every organization. Marriott acknowledging Starwood guest reservation database security incident exemplifies the precarious situation pertaining to cybersecurity today.  

The list of post data breach initiatives is a laborious task. Right from analyzing how it happened to what all was stolen to data breach disclosure and informing customers to implementing security measures after the attack, it is a rough ride for many.

One question that organizations need to dig deep into – how to go for end-to-end protection for the sensitive data an organization has and how to prevent a data breach? Even as one might think over whether data could ever be 100% secure, organizations can’t halt and have to assess how to inch closer to it or bridge a possible loophole? For instance, what’s the weakest point that hackers can go for? Common ways are malware infiltration and phishing.

Stronger protection

This year’s list of impacted airlines includes British Airways and Cathay.

Be it for Marriott/ Starwood or any travel company, what is at stake is possibly what an organization is all about – customers and their data. According to Marriott, the database:

“…contains information on up to approximately 500 million guests who made a reservation at a Starwood property.  For approximately 327 million of these guests, the information includes some combination of name, mailing address, phone number, email address, passport number, Starwood Preferred Guest (“SPG”) account information, date of birth, gender, arrival and departure information, reservation date, and communication preferences.  For some, the information also includes payment card numbers and payment card expiration dates, but the payment card numbers were encrypted using Advanced Encryption Standard encryption (AES-128).  There are two components needed to decrypt the payment card numbers, and at this point, Marriott has not been able to rule out the possibility that both were taken.”

Encryption for personal information

Clearly the days of relying on a simple encryption method are over. In encryption, data is hidden using a coding system that swaps one number or letter for a dissimilar one using a refined encryption algorithm. Encryption of personal data is must and this should span at all possible points where it exists. For this, be aware of where data resides and evaluate cloud settings, big data as well as web storehouses, file systems databases and virtualization implementations.

Companies need to assess latest developments pertaining to database and file encryption.

·          It is imperative to assess what field-level encryption stands for, and once data is encrypted, how systems in a company’s architecture only end up viewing the Ciphertext (it is also known as encrypted or encoded information because it contains a form of the original plaintext that is unreadable by a human or computer without the proper cipher to decrypt it).

·          Also, cybersecurity-savvy organizations are looking at automating encryption deployment and management. Specialists point out that data needs to be encrypted even when it is processed by databases or cached in memory. This can be a critical step as it also cuts down on the risk of access to data owing to the staff’s credentials getting compromised, as data would only be available via authentic applications. (As explained by Microsoft Azure, the state of data at “rest” refers to all information storage objects, containers, and types that exist statically on physical media, whether magnetic or optical disk; in “transit” when data is being transferred between components, locations, or programs). A key is to encrypt application data sent to the database and decrypt query results when it is returned by the database to the application.

Other measures

The role of tokenization, too, needs to be looked at. Even if tokens were to be hacked, it promises to shield credit card numbers or any other critical customer data as none of it would be available for access.

Organizations aren’t only looking at shielding sensitive data, but also to meet regulatory or compliance responsibilities that entail implementation of precise key management controls. It is important to focus on key management process and plan for control of keys that access and encrypt data.

Other than encryption, do plan for a robust user management system. Put in place an incisive access control-mechanism to ensure that only authorized accounts and processes can view the data. Also, gear up for supervision of authorized accounts accessing data, to make sure the same have not been compromised.

The issue of breaches happening due to stolen and/or weak passwords can’t be ignored, too.

The approach needs to be sophisticated. Focus on developing capabilities that can stop attackers at each step of the way to help prevent the theft of data in a breach. Other than investment in cybersecurity technologies, organizations have to hire and retain skilled personnel to form a robust end-to-end protection strategy for sensitive data. Otherwise, it could end up being yet another horrendous data breach story.

 

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Chargeback guarantee + higher approval rates – is it for real?

First Published on 23rd November, 2018

Ai Editorial: As more fraud-related solutions get introduced, the promise of protection against chargebacks is getting stronger. What needs to be evaluated before opting for the same, probes Ritesh Gupta

 

Managing fraud liability and availing chargeback (a forced transaction reversal initiated by the cardholder’s bank) guarantee on every transaction approved at first go comes across as an attractive option when considered from a merchant’s perspective.

Whether there can be 100% prevention of chargebacks remains an interesting discussion, still merchants such as airlines have to work on a risk mitigation plan to cut down on the same.

According to Chargebacks911, chargebacks are caused by criminal fraud (1-10%), friendly fraud (50-80%), or merchant error (20-40%).

Liability shift 

As more fraud-related solutions get introduced, the promise of protection against chargebacks is getting stronger.

But is there any hidden factor that needs to be considered? What needs to be evaluated before opting for the same?

One of the factors is cost vs. chargeback protection.

“Some fraud solutions in the market today offer a guaranteed chargeback protection, which means that they will take financial responsibility for any approved order that turns out to be fraudulent. This shifts the liability away from merchants and onto these fraud specialists,” says Justin Lie, CashShield’s CEO. “However, not all e-commerce merchants will choose to take up the chargeback protection service, depending on their existing chargeback rates and business goals. For example, some solutions factor in the chargeback protection by increasing the cost of service, and a merchant with low chargeback rates may consider their fraud cost lower than the cost of deploying a chargeback protected fraud serviced.”

The cost of the chargeback protected service will be one of the important considerations - if the merchant ends up losing more on cost with the liability shift, then perhaps the merchant would be better off without the chargeback protection.

Second factor is the risk appetite of an e-commerce organization.

A shift in liability might also mean that the merchant would be open to accepting more risk, and therefore more fraud. With that in mind, the travel merchant must consider their risk appetite, whether or not accepting more risk is possible, or if their main goal is to minimize fraud as much as possible. At the same time, fraud rates must still be kept at an acceptable level and not be left too high, or the merchant may be left with warnings and suspensions from card issuers.

Focus shouldn’t be only on “guarantee” 

One can’t ignore the significance of accurately detecting fraud attempts and stopping fraudsters from succeeding in whatever they intend to do.

As specialists at Chargebacks911 point out, merchant errors which can be rather simple and inadvertent need to be curbed. For friendly fraud, a key option for merchant is to strategically argue unlawful chargebacks when they're issued. Each chargeback dispute conveys a powerful message to the issuing bank, asserts Chargebacks911. It also points out by doing so merchants end up restoring their innocence and also improving their association rapport with the issuer. Eventually a merchant freed of any apparent fault are subjected to lesser friendly fraud chargebacks.

In case of criminal fraud, the blend of machine learning with human forensics needs to deliver.

Deploying a multi-disciplinary approach combining different technologies - both supervised and unsupervised machine learning -  would better equip merchants to deal with fraud management. Unsupervised machine learning can be used to learn on the fly and identify fraudulent patterns even without having been trained with historical data, i.e. able to identify unknown fraud attacks. 

Machine learning systems are meant to be an improvement from rule-based systems, to reduce reliance on hard rules and to filter out fraud while passing more genuine users. However, machine learning systems only provide probability scores - or fraud scores - and would still require a team of manual reviewers to make sense of the score and thereafter a decision to pass or reject a transaction.

 

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Biggest question in risk management in payments - are shoppers happy?

First Published on 19th November, 2018

Ai Editorial: Data breaches/ stolen financial information, fraudulent transactions, account takeover (ATO)…merchants have to be at the top of their game to combat commerce-related fraud and at the same time ensure their buyers remain happy with their shopping experience, writes Ai’s Ritesh Gupta   

 

The situation is trickier than ever. Validating a payment could be about running through a gamut of data points and returning with a decision in fraction of a second to optimize a risk strategy.

Payment and fraud executives have to be crafty enough to ensure that genuine customers aren’t denied an opportunity to complete a transaction or even face hiccups with added friction. At the same time, merchants can’t afford to be a victim of fraud owing to weak authentication or fraud prevention mechanism. A specialist like Braintree underlines that simplicity is synonymous with seamlessness. So the checkout process needs to be fluent, but the experience equation isn’t complete without identifying “real” buyers and cutting down on fraudulent transactions.   

Balancing act

So what does balance denote? It’s about being proficient in validating a buyer and such verification shouldn’t interrupt the manner in which they interact and transact with a business.

Merchants need to look at new regulations, what sort of action is required and its impact on the user experience, and also the flexibility of consumes when it comes to additional measures that are being taken for authentication.

In this context, Visa in the recent past initiated new offerings.

For instance, Visa chose to help merchants in following the new European requirements for Strong Customer Authentication (SCA) that come into force in September next year. One way to differentiate between transactions is the risk associated with them. Visa is helping various stakeholders, including merchants, in spotting transactions in real-time that are at low risk and working out a quick checkout for such transactions. Visa asserts that the team ensures a buy is evaluated on a variety of factors such as location, merchant type, previous purchases, transaction size etc. If on one hand, it is closely looking at adding convenient options for any additional step that prolongs the check-out process, on the other, Visa is also capitalizing on the evolution of the consumer technology to simplify authentication. One such move is enabling banks to count on biometric authentication into their respective apps so that users can avail their connected device and use their fingerprint, their voice their face to finish a transaction. The industry is also weighing ongoing improvements, for instance, EMV 3-D Secure Specifications, for a real-time, secure, information-sharing pipeline to authenticate buyers without adding any friction in the buying process.

Other than tools, global merchants like airlines, also need to dig deeper and understand the peculiarities of each market before taking steps to streamline the buying process.

In its 2018 Global Fraud and Identity Report, Experian shared critical findings (based on 11 countries surveyed):

·          Consumer tolerance for friction for the sake of security varies with reference to various nations.

·          This is a dynamic place and merchants need to be vigilant of the changing scenario. Considering the variance in tolerance for friction, one solution is hard to finalize. Also, since it is “a moving target”, companies need to adjust especially if buyers’ tolerance for friction could lessen over time.

Being smarter 

Merchants need to understand the limitations of existing measures, and also leverage the prowess of data-driven, artificial-intelligence powered offerings for combatting fraud. Rules-based systems are in general reactive and probabilistic solutions, which is why they are unable to prevent fraud before it happens. Rather than using a blanket rule that forces every user to login with 2FA, real-time surveillance can be used to assess logins in the background, and only logins with borderline risks expected to go through 2FA.

Merchants should still develop their own fraud tools that are able to tap on their own sources of data for greater efficiency and more accurate detection of fraud. Real-time machine learning can help against blanket blacklists and whitelists by focusing on the customer’s behaviour instead. It works with real-time live data collected on the merchant’s website, where the system trains itself with each incoming transactions to identify fraud patterns instead.

 

Check upcoming Ai Conferences dates or

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Here’s why account takeovers are set to become a bigger headache

First Published on 8th November, 2018

Ai Editorial: Account takeovers (ATO) are shaking e-commerce players in many ways, including in the loyalty space. For instance, post an ATO orders can be made with the genuine card-on-file or stored credit (reward points or miles), writes Ai’s Ritesh Gupta 

 

Retailers, including travel e-commerce players, are looking at combating the increasing threat of account takeover (ATO) attacks.

As the number of data breaches is going up, they are being linked to the surge in ATO attacks. This is because these breaches supply a treasure trove of information of login credentials, passwords, and personal information.

Here is how fraudsters are trying to make sense of what they are stealing: Data breaches can result in compromised login credentials. Post this fraudsters tend to test whether these credentials work on other sites or not. With one password for multiple accounts being a common practice, the threat of danger is unimaginable! Since testing credentials this way can be a laborious task, fraudsters use bots to automate the testing process. Once fraudsters have found credentials that work, they can either commit the fraud, or sell them on the dark web. According to Riskified, usually fraudsters have specialized roles: fraudster A is the expert at data breaches, and he’ll monetize the stolen credentials by selling them to fraudster B, who is the expert at loyalty fraud. Cybercriminals purchase these stolen credentials from the dark web, and thereafter launch coordinated fraud attacks for hostile ATOs, or to create spam accounts with real genuine identities.

Why even a bigger headache? 

In case a successful data breach or an ATO attack happens, merchants can find themselves in an obnoxious situation. As explained by Sift Science, this is because stored payment methods make it easier for fraudsters to shop, fraudsters can redeem miles and points that sit in unsecure accounts, personable information is lost and then merchants also have to grapple with the issue of restoring accounts after a takeover incident. Spam accounts are useful for fraudsters to abuse promotional codes, which is another pain point for merchants.

A couple of reasons by ATO can become even a bigger headache:

·         ATO can also occur under other circumstances such as when competitors suffer a data breach. Given that most people tend to use the same credentials with multiple merchants, fraudsters will test stolen credentials across multiple websites. This means that an enterprise’s accounts can also be compromised once fraudsters get hold of their competitors’ user data.

·          It also needs to be considered that enterprises are starting to build ecosystems where a single account can be used to access multiple services, increasing the value of accounts and further compounding the problem of account takeovers. Accounts are becoming increasingly valuable, due to the amount of information and/or services tied to a single login, and considering that most enterprises have yet to deploy sophisticated fraud management techniques to detect fraudulent account logins, accounts have become the new gold for fraudsters today. A case to examine is Amazon, where one single account may be used to access multiple services including Amazon Prime, Alexa, cloud storage, music streaming and more. Once a single account is compromised, it would be difficult to have damage control on all possible endpoints that could benefit the fraudster. For instance, the fraudster could have access to the card-on-file to make purchases, or have access to the user’s information, or worse, in the case of IoT (e.g. Alexa), spy on the users in their homes.

What to consider? 

Some of the top issues on the agenda of airlines as of today are - how to prevent fraudsters from accessing travellers' legitimate account? How to combat an ATO attack at the point of sale, and declining the order?

Companies acknowledge what's at stake - their reputation, messed up loyalty accounts, a customer's private information etc. A majority of fraud review operations are reluctant to decline orders coming from a logged in account. This is because the risk of offending a good customer is so high and the fear of a poor customer experience makes it a delicate issue. As pointed out by Riskified, a major aspect of preventing fraudsters from succeeding at the point of account login is processing data and making decisions in real-time.

Enabling two-factor authentication (2FA) is one option. Educating consumers to use strong passwords and securing their devices is also important. Notifications about suspicious activity, too, need to be considered. Still travel e-commerce companies need to dig deep. As recently shared by CashShield, organizations tend to rely on 2FA for account protection, which can be overcome by fraudsters with deceptive tactics, such as SMS phishing to trick users into giving up their 2FA reset codes; it is also not uncommon for fraudsters to intercept the confirmation SMS messages, proving that 2FA is not sufficient to prevent fraudulent account takeovers.

As for the role of a merchant, they need to go for stringent security protocols in storing and encrypting their data, to curtail the loss in case of a data breach. They can also attempt to lessen the harm by guaranteeing that the stolen data cannot be used. According to CashShield, one way to achieve this is to deploy real-time active surveillance on every login to filter out potential threats and prevent attackers from gaining unauthorized access to accounts.

 

For Ai’s Events, check - www.aieventdates.com

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Machine learning and fraud – “scores” not important; results matter

First Published on 16th October, 2018

Ai Editorial: There are key pointers – denial rates, false positives and fraudulent transactions – that underline the performance of any machine learning technique in fraud prevention. As for what is the utility of scores, they are not important; results matter, writes Ai’s Ritesh Gupta  

 

It is intriguing to understand how machine learning works – working on data, variables etc., and how is precise model worked out and refined to control fraud, be it for related to a payment, data breach or account takeover.

The machine learning system starts with a basic model which is trained and improved with datasets over time. It is important to pre-process the data. To improve the efficiency and accuracy of the system, the data can be pre-processed with data slicing and augmentation and be cleaned sufficiently before it is used to train the model.

Making it work to control fraud

In the case of a fraud solution, the system will be given training sets consisting of a given set of known fraudulent transactions and known non-fraudulent transactions, so that the system will learn to differentiate and filter away fraudulent transactions, says Justin Lie, CashShield’s CEO. 

Considering that various industries have differing levels of risk and exposure to fraud, the data collected from different industries may be customized. For instance, some data sets that may be collected from an airline merchant (and no other industry) would include: flight boarding times, whether the customer chooses to add a meal, whether the customer has an existing loyalty membership or whether seat preferences have been added.

A few algorithms modelled from the training sets will be put to the test with real life data, and thereafter, the algorithm with the least error will be chosen as the best algorithm. The amount of data and how relevant the data was used in deriving the algorithm will affect its accuracy. Over time, the algorithm must be constantly trained with data, especially with new data so that the margin of error can be minimized and inaccurately classified transactions (fraudulent as non-fraudulent and non-fraudulent as fraudulent) will be corrected.

Significance of “score” associated with machine learning 

There are key pointers – denial rates, false positives and fraudulent transactions – that underline the performance of any machine learning technique. As for what is the utility of scores, Lie says scores are not important; results matter.

“Most merchants would aim to increase their transactions and reduce their fraud, and the performance of any machine learning technique should be evaluated based on whether this goal can be achieved. Nevertheless, it is important to note that each merchant would have differing goals with respect to fraud; for some, raising acceptance rates and growing aggressively is most important, while for some others, minimizing fraud rates down to zero is the most important KPI,” says Lie.  

“With minimal risk, it is likely that overly strict filters are put in place and many genuine users have been blocked at the expense of lowering fraud rates. Therefore, the performance of the fraud solution would depend on the goals of the merchant. For example, taking in more risk may increase fraud rates slightly, but also lower false positives and rejection rates.”

Commenting on the significance of scores in terms of performance in controlling fraud and letting legitimate transactions go through, Lie said most fraud solutions on the market would be able to automate a good bulk of the transactions based on the score; extremely low scores will be rejected automatically and extremely good scores will be accepted. However, for the borderline transactions, a team of manual reviewers is required to make sense of the score. Generally, some guidelines will be given to the manual review team to look for further clues based on the data collected, and some working experience will be used, but most of the time the manual review team is relying on their gut feeling, which is affected by a risk-averse outlook to reject potentially genuine transactions to prevent fraud rather than to risk having passed a fraudulent transactions.

Therefore, fraud scores can only help a merchant this much, but ultimately, the fraud score is not the be all and end all in identifying fraud.

Human intelligence counts

Machine learning models would only still provide merchants with only a fraud score; to make sense of the score, fraud solutions or merchants would still need to rely on humans to make a decision.

“The problem here, is that humans are often risk-averse and would reject borderline risky transactions for fear that it could be fraudulent, and end up blocking more genuine customers than expected,” said Lie.

As such, a multi-disciplinary approach combining machine learning and other techniques is important to improve the efficiency and quality of the fraud detection process.

 

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: How agility is playing its part in payment optimization at KLM?

Ai Editorial: A tailored payment infrastructure and the structuring of team internally, where multiple teams working in sync within an agile environment, paves way for payment optimization at KLM, writes Ai’s Ritesh Gupta. He spoke to Maarten Rooijers, Senior Manager Payments, KLM in Phuket recently.  

 

In an era where the number of ways in which a customer can pay has risen tremendously, facilitating such wide variety of payment methods can be an arduous task. A merchant today is possibly expected to facilitate a transaction via every point of interaction.

In this context, an organization’s of KLM stature has led the way in embracing new technology and payment methods in a swift manner.

Consumers are indulging in technology and making the most of new devices to simplify what they wish to do. They are shopping for various products, including travel, through “voice”. Seeking Alexa and Google Assistant’s help in one’s native language is becoming a norm, and brands like KLM are responding to this trend. The airline is allowing users to search flights by giving spoken instructions. Once a suitable flight is identified, a link to the KLM's site is provided to complete the transaction. KLM’s Blue Bot is based on artificial intelligence, which is linked to a combination of KLM and external tech.

So how to optimize the payment experience by balancing the cost vs. revenue analysis or assessing the intangible value gained from any initiative? 

“There is a need to evaluate whether the new technology (or any payment method) would result in additional value for the customer as well as the merchant. It is the customer who decides how they wish to pay,” says Maarten Rooijers, Senior Manager Payments, KLM. Rooijers was recently in Phuket for Ai’s ATPS, where he explained the evolution of KLM's online payment strategy (alternative forms of payment online, leveraging social media, multi-currency pricing, roll out of mobile wallets, demo of payment via WeChat etc.). 

“Some of the new options to pay are being propelled by innovation in this industry. Some are also being facilitated by social media,” mentioned Rooijers.

 

The Payment team works closely together with KLM’s Social Media team which is also part of KLM’s Digital department. “Apart from the Payment team coming with initiatives to add new Payment options, it is sometimes a combination of initiatives coming from our establishments globally and requests from our SM team. It is not just about adding Google Pay, Apple Pay, Alipay etc., but it also about making the booking process easier. As for the payment team, we chose to standardize the process. So rather than having a payment functionality in each and every front-end, it was decided to set up an independent payment platform or a payment engine. It is connected to “internal” customers/ front-end for payments,” explained Rooijers. 

Infrastructure + Agility 

The payment infrastructure and internal alignment paves way for payment optimization at KLM.

“Internally, we started working via structuring or a framework like Scrum (to embrace agility). The number of product teams within the KLM digital is quite big. There are multiple teams working in sync within this agile environment, involving the front-end, back-end API teams, payments team…looking at implementing new projects/ features.” shared Rooijers. 

 

In the agile set up, how often does the payments team interact with the other teams?

“It’s almost daily,” mentioned Rooijers.

The Payment system is called EPASS (Electronic Payment and Settlement System). As for how the team manages challenges pertaining to introducing a new payment option, for example, Alipay or WeChat Pay, Rooijers shared that the EPASS layer is what that maintains liaising between the internal applications, mostly front-ends where customers book their trip or buy their ancillaries, and the external partners/ vendors – payment service provider (PSPs), gateways etc. “We are working with a number of PSPs and acquirers in order to be able to have a global offering. So when it comes to facilitating viable payment options from a new market or a region, rather than directly working with a new PSP, we prefer to route them to our existing PSP and work accordingly. Making a connection to a new PSP tends to be costly and needs resources.  We therefore have selected PSP’s with a wide global coverage. If there are any new relevant Payment options emerging not initially supported by our contracted PSP’s, we request our PSP’s to start offering these or, alternatively, partner with the local or regional PSP.”

“Having said that”, adds Rooijers, “we always will keep our eyes and ears open for other players in the Payment landscape. The market is changing constantly and rapidly and we want to continue to offer relevant payment options globally at the right costs and according the most efficient process.”

KLM is offering 80 alternative payment options, and 10 of them are from the Asia Pacific region.

 

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Working out an apt infrastructure for payment optimization

First Published on 21st September, 2018

Ai Editorial: Airlines intend to take steps to control their payments strategy holistically. This gives them visibility into  payments in all their sales channels and in all countries, allowing them to identify inefficiencies, writes Ai's Ritesh Gupta

 

What can pave way for optimization of payment-related experience? It is a vital question, especially for travel merchants with cross-border operations.

Businesses can't wait for long to embrace a new payment method. Every minute detail is being scrutinized as far as removing friction from the shopping process is concerned. And if a customer abandons the shopping cart owing to inflexibility in payment-related options, then the likes of airlines aren't doing any good to their business.

Roadblocks to innovation

According to a recent study by Amadeus' payment business, travel e-commerce companies are pulled back by many factors when it comes to innovation in this space. The list includes consumer data security, credit card data security, fraud losses, complexity of existing payment systems etc.

So how to combat such major roadblocks to payments innovation?

"Airlines looking to innovate while at the same time keeping customer data safe should first carry out an audit of all the systems in their infrastructure which are storing or using that data," recommends Klein Wang, Head of Merchant Services APAC, Amadeus’ payments business.

Wang further added, "Then look to limit – or even eliminate – that data from their systems using techniques like tokenisation. When looking for a tokenisation provider, it is critical to find a provider whose tokens are compatible with all those systems which will need to use and interact with them."

Simplifying infrastructure

There tend to be issues with payment-related infrastructure and the role of partners that facilitate a payment. For instance, not all global payment processors handle acquiring banks and routing the same way. "Airlines work with a number of different acquiring banks and other related service providers. To simplify their payment infrastructure, airlines should look for payments providers which can give them access, control and visibility over their payments infrastructure in a single place. A single source of data and a single entry point to monitor and identify payments from across all their providers," mentioned Wang.

Also, talking of payment acceptance, it is vital for airlines to minimize costs with a more fluid workflow. For instance, what needs to be considered to route a payment to an appropriate acquiring bank? What does the study by Amadeus suggest? "It’s important for airlines to look at the total cost of their payments infrastructure, not just the direct acquiring cost. And many of them are doing that – although not all. With that in mind, the greatest opportunity for airlines today is in reducing the indirect costs of accepting payments caused by manual processes and inefficiencies which are the result of the complexity of managing payments across many different providers in different sales channels and markets," said Wang.

Alternate payment methods

According to Amadeus, many airlines have been adding new payment methods and today accept on average between six and ten different payment methods.

There are two challenges for airlines adding more payment methods: user experience and reporting.

"From a user experience perspective airlines need to find a way to offer the right payment method to the right customer without providing a bewildering list of different payment methods. From a reporting perspective, each new method of payment has a different reporting format so airlines need to work with a partner which can help to harmonise all these different formats so they can get a single view of all payments from all payment methods," said Wang.

Other than most popular or new options, airlines also need to work out a local checkout experience and at the same time being in a position to curb possible criminal fraud.

Wang mentioned that there is always a trade-off between verifying that a payment method is being correctly used by its owner on the one hand and creating an easy-to-use and seamless check-out experience.

"Using a fraud management tool to check the details for signs of potential fraud – in a way that is transparent to the end user – can really help here. Combining such techniques with the application of stronger authentication practises, 3D Secure for example, when the fraud management tool suggests dubious transactions, is a good way to balance consumer experience with effective fraud control," said Wang.

Airlines need to sort payment infrastructure related issues. In fact, as Wang says, for airlines looking at passengers from Asia, these problems even get more complex. "Many airlines are still not offering book and pay functionality on their mobile apps – in many markets, however, mobile is the default. In China, for example, 73% of all e-commerce purchases by value are made on a mobile device (Kleiner Perkins) and mobile payment volume grew 116% in 2017 over the previous year," said Wang. Asia has more diverse local payment practices, more local regulations and more payment providers to work with.  

 

Follow Ai on Twitter: @Ai_Connects_Us

 

Ai Editorial: Use of app cloners and machine learning rises in mobile commerce fraud

First Published on 17th September, 2018

Ai Editorial: Considering the growth of mobile commerce, it is imperative for travel merchants to assess how fraudsters are crafting new techniques and tools for executing fraud through mobile devices, writes Ai’s Ritesh Gupta

 

Merchants can’t ignore mobile commerce. In fact, it doesn’t come as a surprise to see the way travel merchants are enhancing their apps with digital features and capabilities. The consumer is mobile and interconnected across multiple devices, shopping and switching from one to the other.

Shopping via mobile devices is flourishing. The growth rate is faster in certain countries, with China leading the way (m-commerce reached 93% of all e-commerce sales last year in China). In the U. S., an estimated 30% of all online sales are made on mobile devices. In Europe, transactions on mobile and desktop are estimated to be split 50-50. But is there any threat that is being ignored by travel merchants as the world of retailing is now mobile-first in most parts of the world?

“A new mobile commerce-related fraud technique that we found has emerged is the use of app cloners and machine learning techniques to create synthetic device identities,” says Justin Lie, CashShield’s CEO.

Mobile and synthetic identity fraud

Synthetic-identity fraud has already been described as the fastest-growing forms of identity theft in the U.S., according to the Department of Justice, as reported by CNBC.com in June this year. This kind of theft is working out a false identity using either fabricated or valid elements, such as a Social Security number, name, address, date of birth etc. As we highlighted in one of our articles this year, it often is not identified as fraud and the crime can go undetected for an indefinite period.

Even cloning has proven to be an issue for those who have been trying to control mobile device fraud.

For instance, a SIM swap attack in which a mobile number is hacked remains a problem. The fraudster takes control of a legitimate mobile user’s text messages, calls etc. Then login credentials are obtained through social engineering, phishing, an infected downloaded app etc. Mobile apps being reverse engineered and adding malicious code, too, has been around for a while.

As for how app cloners are being used in a malicious way, Lie shared, “Using app cloners, fraudsters can masquerade as multiple users to trick systems since the different transactions or logins will be detected as unique devices,” shared Lie.

Swift response 

In one of its recent blog posts regarding mobile bank fraud, Microsoft Azure stressed on the significance of latency and response time when it comes to fraud detection. The team mentioned that the time taken by a bank to respond to an illegitimate transaction “translates directly to how much financial loss can be prevented”. The response time window or detection needs to happen in mere two seconds.

“This means less than two seconds to process an incoming mobile activity, build a behavioral profile, evaluate the transaction for fraud, and determine if an action needs to be taken,” recommended the blog.

In this context, as Lie also stated, machine learning can be used to identify such new forms on fraud.

“For instance, whether or not an app cloner has been used may be collected and analyzed as one of the data points, in addition to the other various data points that will be analyzed to identify hidden patterns between the same transactions by the same fraudster,” mentioned Lie.

Making the most of signals

As for what signals are being considered for transactions coming from mobile when it comes to previous purchases and also pattern recognition for possible fraud in the future/ unknown attacks, Lie indicated that rather than focus on the signals for fraud, the company often tends to focus on signals that may point to positive genuine behavior of the user making the transaction. For example, if the user chooses to connect with social media, that is more likely genuine as most fraudsters who want a quick exit would not bother connecting with social media in making transactions.

“However, such signals are not necessarily accurate in pointing out fraud - a more sophisticated fraudster might put in more effort to imitate a genuine user, and connect with social media just to trick the system. Therefore, what is more important in identifying unknown attacks would still be to use real-time pattern recognition to draw patterns between incoming transactions, and identify coordinated fraud patterns,” added Lie.

 

Follow Ai on Twitter: @Ai_Connects_Us

Editorials

  • Ai Editorial: 5 reasons why data is set to play an integral part in combatting fraud +

    First Published on 14th December, 2018 Ai Editorial: Data is going to be a key weapon in the arsenal of airlines as the industry attempts to fight emerging fraud threats, Read More
  • Ai Editorial: From being NDC-ready to implementing it - how IATA is focusing on it? +

    First Published on 7th December, 2018 Ai Editorial: IATA has been evaluating the performance of NDC - what is missing to attain 20% sales-target that is going to be powered by Read More
  • Ai Editorial: Why even one data breach today is enough to shake a consumer? +

    First Published on 4th December, 2018 Imagine receiving an email early in the morning, stating that your personal data has “possibly” been compromised. It’s disturbing. But the agony doesn’t end Read More
  • 1
  • 2
  • 3
  • 4
  • 5