a.i. led security€¦ · spectre is harder to exploit than meltdown, but it is also harder to...
TRANSCRIPT
A.I. Led Security Steve Grobman, McAfee
Steve GrobmanSVP & Chief Technology Officer
McAfee
New Attack Methodologies Create New Challenges
Artificial Intelligence
Machine Learning
Deep Learning
Ideally suited for a wide range of cyber security defense use cases…
Detection rate: 88.65%False Positive rate: 0.00%
Threat Intelligence driven security
Detection rate: 97.42%False Positive rate: 9.67%
Machine Learning driven security
Detection rate: 98.86%False Positive rate: 0.38%
Threat Intelligence + M.L. driven security
A.I./ML
ThreatIntelligence
Adversarial Machine Learning
(~100%) (~50%) (<1%)
Artificial Intelligence “Black Box Duality”
• Detect Previously Unseen Attacks
• Non-Deterministic
• Often Unable to Explain “Rationale for the conclusion”
• False Positives and False Negatives are inherent
In comparison to traditional risk measurement; the focus is on probability instead of possibility
Business Function 5
Business Function 4
Business Function 2
Business Function 3
Business Function 1
One final thought
NIST physicists used three beryllium ions to demonstrate a crucial step in a procedure that could enable future quantum computers to break today's most commonly used encryption code
…We must act now!!