2019
DOI: 10.1038/s42256-019-0109-1
|View full text |Cite
|
Sign up to set email alerts
|

Trusting artificial intelligence in cybersecurity is a double-edged sword

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

0
54
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
5
4

Relationship

2
7

Authors

Journals

citations
Cited by 105 publications
(59 citation statements)
references
References 13 publications
0
54
0
Order By: Relevance
“…Especially during the pandemic, a large part of Australian society was intending to move towards digital transformation. Due to the cyber-attack boom in 2020 in Australia, the discussions on cybersecurity and the ethical concerns associated with using AI technologies become highly prominent [62,63]. Australian researchers have also highlighted the importance of understanding the loopholes in the present AI systems [64].…”
Section: Discussionmentioning
confidence: 99%
“…Especially during the pandemic, a large part of Australian society was intending to move towards digital transformation. Due to the cyber-attack boom in 2020 in Australia, the discussions on cybersecurity and the ethical concerns associated with using AI technologies become highly prominent [62,63]. Australian researchers have also highlighted the importance of understanding the loopholes in the present AI systems [64].…”
Section: Discussionmentioning
confidence: 99%
“…A connected city can emerge as a hacker's playground, as already seen via the "Krack attack" on Wi-Fi, a series of hacks on Amazon's Ring home security camera, CCTV camera hacks, and the list goes on [113]. While the use of AI to protect our cyberspace has a substantial scope for improvement, more integration could mean more vulnerability due to lack of technical capabilities, lack of understanding, lack of prioritizing cybersecurity, budget constraints, technological dependencies on other countries, the territorial reach of laws and regulations, lack of trust and transparency-all of which require continuous synchronized adaptation, based on circumstantial requirements [114]. Schneier portrays "a digital nightmare of hyper-connectivity" due to cyber vulnerability and the ways that policies can be designed to tackle them [115].…”
Section: Discussionmentioning
confidence: 99%
“…Adversarial training between AI systems can help to improve their robustness as well as facilitate the identification of vulnerabilities of the system. The more we adopt an "adversarial" and "always-on" strategy, the safer the AI applications become [16].  Parallel and dynamic monitoring: The learning abilities of the targeted systems require some form of constant monitoring during deployment.…”
Section: Applications Of Ai In Cybersecuritymentioning
confidence: 99%
“…Monitoring is necessary to ensure that divergence between the expected and actual behavior of a system is captured and addressed adequately. To do so, providers of AI systems should maintain a clone system as a control system, which serves as the benchmark against which the behavior of the original system is assessed [16].…”
Section: Applications Of Ai In Cybersecuritymentioning
confidence: 99%