Sunday, February 16, 2025

A Fine Line Must Be Drawn In Generative AI Usage: The Banking Example

 


One common question that I get asked from time to time is what do Cyberattackers like to prey on?  In other words, who do they like to target?  To be honest, about anything and anybody can be a prime target.  But it all comes down to one key motivating factor:  MONEY, AND LOTS OF IT. 

Wherever there is a backdoor is open and $$$ is easy to smell, the Cyberattacker will make its prey.  It can happen in a lot of diverse ways, such as Social Engineering, Phishing (Business Email Compromise is a big one here), finding vulnerabilities in a web application, etc.

But one thing I can answer for sure is that an industry which is heavily targeted is the banking  one.  After all, once the Cyberattacker has access to the login info of the victim, all heck can break loose.  For example, they can initiate a fake transaction, open a fake debit card, or just do things the old-fashioned way:  just steal whatever money is in the victim’s account.

In response to this, most of the financial institutions based here in the United States have done an excellent job implementing safeguards to protect their customers.  I can even vouch for this for myself.  One time, I got a letter from my bank stating that my debit card got hacked into. 

I never even used it, but the moment they got whiff of a potential, fraudulent transaction, they cancelled it immediately.  Then one time, I logged into my checking account from my iPhone (which I hardly ever do), the bank blocked my access later, because a different IP address was detected.

But another area in banking which needs more attention paid to is that of the mobile apps that they create and deploy for their customers.  Consider these stats:

*Fraudulent activity will exceed $40 billion by 2027, which is a staggering 32% increase.

*Banking as a Service will also witness a 20% increase in attacks.

(SOURCE:  How Banks Can Adapt to the Rising Threat of Financial Crime)

In fact, the mobile app can be viewed as Banking as a Service tool.  After all, you can download it from an app store, such as Google or Apple.  In these cases, one of the easiest ways for the Cyberattacker to get into is to try to find a backdoor in the source code, especially in the API. 

As I have written before, many software developers use ones that are open sourced primarily because they are free to download and use, with no licensing fees involved.  Also, there are plenty of forums online in which help, and resources are available.

But the software developers who make use of these kinds of APIs do not check them to make sure that they have been updated.  Because of this, many backdoors can be left open for easy penetration by the Cyberattacker.  From here, they can manipulate the mobile app or even heist the source code to create a fake, this tricking and luring in their victims.

So how can a bank avoid this situation.  In a theoretical sense, the easy answer is that they should use their own IT Department to create it.  But, this can be a costly proposition, so many banks choose to outsource the development of it, in the name of saving money. 

While this can be a good thing, it also poses grave risks as well.  For example, what if they have hired a web development team, such as in India, and they are not properly vetted?

In this regard, the banks must take the vetting process very seriously.  They need to make sure that whoever they hire must meet strict security requirements that are at least on par or even greater than what the bank has in place. 

Further, the right controls must be put in place, in case any customer information and/or data is given for testing purposes.  In fact, the bank should take the initiative and responsibility to create a set of best practices and standards for their vetting process.

Another avenue that banks are looking at to further protect they are as a Service offerings is the use of Generative AI.  One of the best ways that this has been used is to quickly detect any form of abnormal behavior that falls out of the baseline profile of the customer.  

Once this has been captured, the Generative AI model will trigger the account to be blocked almost immediately.  Generative AI is also great when it comes to halting a wire transfer that looks fishy, such as the in the case of a Business Email Compromise Attack.

But with the good comes the bad.  For instance, a Cyberattacker can easily heist one of these models and modify it in a way that it will not detect fraudulent activity for a certain period of time.  Or worse yet, they can not only create a fake website,  but they can also Generative AI to create a Deepfake, which is a replication of a real person. 

They can use this to create a Digital Personality that the customer can interact with, but Social Engineering can be embedded here, so that a trusting dialog can be developed.  Once this has come to fruition, Digital Personality can then be manipulated to prey upon the vulnerable state of mind of the customer and con them into giving out their personal information and data.

My Thoughts on This:

IMHO, banks, no matter what their size or their geographic location is, there must be a fine line drawn as to how much Generative AI should be used.  Perhaps creating a set of best standards and practices would be great here, as to where it can and cannot be used.

In the end, it is extremely easy to get swept away by the glamor that Generative AI brings to the table,  but it is especially important to keep in mind, as in the case of the banks, that the human side is needed as well.

Back to my example again of my account being blocked.  Suppose the only way that it could be unblocked was by having a conversation with a Digital Person.  But for some reason, no matter how much I tried to convince it that it was me that was trying to log in, it still does not unblock it. 

But luckily after waiting for a few minutes, I was able to reach a real, live customer assistant to whom I explained the situation.  The next second, it was unblocked.

The equation for having a great level of security is to have a balance between technology and the human element.

No comments:

Post a Comment

A Fine Line Must Be Drawn In Generative AI Usage: The Banking Example

  One common question that I get asked from time to time is what do Cyberattackers like to prey on?   In other words, who do they like to ta...