AI-powered cyber threats in South Africa written By Myles Illidge

Although artificial intelligence (AI) currently has more shortcomings than benefits, that won’t remain the case in the long run, writes Professor Basie von Solms.

Von Solms is a research professor specialising in cybersecurity at the Academy for Computer Science and Software Engineering at the University of Johannesburg.

He said that, at the moment, AI-driven products primarily work indirectly, such as a bank using AI technology to analyse data to detect fraud and money laundering.

Von Solms said ongoing research could lead to more sophisticated products directly benefiting the average person.

“The businessperson would like a product that can review their daily schedule and then optimise it to use available time as efficiently as possible — make and cancel meetings and appointments,” he said.

For example, an AI assistant could identify a clash between an important meeting and a golf day and then come up with a believable excuse as to why you can’t attend the meeting.

“Teachers and lecturers want a product that helps with marking and evaluating assignments — that would directly benefit their lives,” says Von Solms.

“Parents want AI to manage and optimise their children’s school and extracurricular activities — ideally for the whole term.”

He noted that one reason products like these aren’t widely available is privacy. For example, business people would have to share a lot of personal and confidential data with an AI assistant.

Parents using an AI tool to optimise their children’s schedules would have to provide the product with all relevant dates, times, and appointments to enable such optimisation.

“While such experimental products may already exist, they are far from ready for everyday use,” said Von Solms.

“On the other hand, there are AI-driven products already in use that can directly and seriously affect our lives negatively in the short term. And here the picture looks bleak.”

According to several research reports, cybercriminals are among the top AI users, and they use them in ways that could impact the average person on a daily basis.

For example, cybercriminals now use AI to bypass and neutralise antivirus software to infiltrate PCs, tablets, smartphones, and other devices.

They also use AI for phishing purposes. Phishing is a social engineering technique through which fraudsters attempt to coerce victims to share sensitive information by imitating a legitimate business, such as a bank.

“Here, the attacker uses various AI techniques to make the message as convincing as possible to the victim, who believes the request is legitimate and real,” said Von Solms.

He added that these phishing attacks have been particularly successful when the attacker contacts victims by phone.

In many cases, fraudsters have a full profile of the potential victim. Using AI, they compile personal and confidential information from various sources.

This could be through your social media profiles, leaked data posted on the dark web, and various other sources.

“With such a comprehensive profile in hand, the attacker pretends to be calling from your bank, usually from the fraud department, and gradually earns your trust by sharing more profile information with you,” said Von Solms.

“Millions of rands are being stolen this way.”

Other ways through which cybercriminals use AI to take advantage of businesses and people include business email compromise attacks, the use of deepfakes, and undressing apps.

Through business email compromise attacks, cybercriminals can intercept invoices and alter them using AI to change the bank account for payment to the attacker’s account.

AI can make these altered invoices so convincing that they are difficult to distinguish from the original.

AI can also enable AI to create convincing deepfakes to deceive victims into falling into a trap.

“Using AI-driven deepfakes — of both voices and videos — victims now receive a voice or video message supposedly from a family member or friend urgently requesting money or help,” said Von Solms.

Undressing apps effectively undress photos to create a naked version of it.

“Determining any positive value of such apps is difficult, but the negative impact is obvious,” adds Von Solms.

source:AI-powered cyber threats in South Africa – MyBroadband