In an era of large-scale cyberattacks, some of which can be bought on the dark web for as little as $10, the security industry is increasingly turning to artificial intelligence to come up with adequate defenses.
The growing complexity of IT systems, the increasing aggression of state actors and the challenge of undereducated digital users are driving the adoption of AI-based cybersecurity technology designed to better forecast, prevent and mitigate digital threats, according to panelists at CogX, an annual artificial intelligence conference in London.
Cath Goulding, head of IT security at Nominet, a registry for U.K. domain names, highlighted a shortage of cybersecurity skills as one reason for the uptake in AI.
"There are not enough people [or] analysts to tackle this problem," Goulding told delegates, adding that the high number of anomalies detected in systems meant that analysts can become "quickly fatigued."
With the constantly evolving nature and attacks becoming more detrimental to businesses, IT security professionals are making greater use of AI through algorithms trained on large datasets of malicious programs in order to learn what to look out for.
Also speaking on the panel, David Atkinson, CEO of cybersecurity firm Senseon, said that by using deep learning to classify malware, particularly as businesses generate unprecedented amounts of data, AI and machine learning are enabling the industry to take a more efficient approach.
"It allows us now to begin modeling normal behavior of users and their devices and detecting outliers within datasets to detect new and novel techniques," Atkinson added.
At the same time, machine learning-based solutions must not operate in a vacuum, particularly as attackers have begun to adopt AI techniques themselves, fueling a wave of new threats, the panelists agreed.
Instead, AI is useful for providing additional horsepower to an existing cybersecurity arsenal in order to improve operational efficiency when it comes to tasks like malware or spam detection.
"Machine learning is excellent for intelligence but you have to be a little more cautious when it comes to pure defense," Goulding said.
This approach means that machine learning's most vital role becomes understanding the characteristics of an uncompromising system, and then flagging any outliers for human review, she explained.
By using AI as merely complementary to your other sources, Goulding concluded it provides greater agility in what has historically been a manual and repetitive procedure.
Additional CogX 2018 coverage:
Artificial Intelligence won't create killer robots, mass disruption