Meet your
speakers
Dr. Emmanuel Tsukerman
Guest speaker
Infosec Skills author
Agenda
● What is ML?
○ It automates and scales
○ It’s flexible
○ It allows innovative attacks
● Deepfake live demo
● Few more ML use cases
● Career outlook
What is machine learning?
Short answer #1:
ML automates and scales
Example: ML for malware detection
5
6
Traditional anti-
viruses (AVs) rely on
signatures and
heuristics
• Signatures: Hashes of files, hashes of
snippets of code; akin to cataloging
• Heuristics: Similarity of source code to
known malicious samples, self-replication,
overwriting files, suspicious network traffic,
etc.
7
Malware is too numerous and evolves too fast
>350,000 new malware created every day (https://www.av-test.org/en/statistics/malware/)
How do I know traditional AVs
don’t keep up?
8
9
ML-based AVs are now industry
standard
10
What is machine learning?
Short answer #2:
Flexibility
Example: Network intrusion detection
12
13
Problem: There is a never-ending
array of methods of attack
14
Illustration of threat diversity
15
DOS: SMURF, NEPTUNE, Back, land, PoD, Teardrop, TCP SYN flooding …
16
Illustration of threat
diversity from
firewalls
● Packet fragmentation: Allows hackers to
evade pattern matching by fragmenting
their packets
● Spoofing source IP address: Hackers
appear like legitimate users to the firewall
● Spoofing source port: Allows hackers to
bypass port-specific rules in the firewall
New attacks developed everyday
17
🙶 It’s humanly impossible
to keep up with this ever-
evolving threat landscape
🙶
ML can keep up with and catch:
● Zero-day attacks
● Anomalous behavior
18
What is machine learning?
Short answer #3:
Innovative attacks
20
You won’t believe what Obama says in this video!
—BuzzFeedVideo
21
This type of crime is
already here
“Fraudsters Used AI to Mimic CEO’s
Voice in Unusual Cybercrime Case”
—Wall Street Journal
Deepfake live demo
Few more illustrative ML
use cases
Hacking CAPTCHA systems
24
CAPTCHA is meant to tell humans and
computers apart, but we can break it using
ML
Smart fuzzing
25
● Fuzzing is about finding inputs to a
program that will cause it to break
● The best fuzzers rely on ML (e.g., genetic
algorithms)
Career outlook
Plenty of great jobs
27
The primary profession for
applying data science to
cybersecurity is “cybersecurity
data scientist”
28
ML is beginning to
touch all aspects of
cybersecurity
Areas influenced:
● Malware
● Intrusion
● Social engineering
● Pentesting
● Data security
Best learn it, because the bad guys are
Dr. Tsukerman’s
Infosec Skills courses
infosecinstitute.com/authors/emmanuel-tsukerman/
Questions?
Everyone gets a free week
of Infosec Skills.
Then it’s just $34/month
infosecinstitute.com/skills
About us
At Infosec, we believe knowledge is the most
powerful tool in the fight against cybercrime. We
provide the best certification and skills
development training for IT and security
professionals, as well as employee security
awareness training and phishing simulations.
infosecinstitute.com
708.689.0131
32

From machine learning to deepfakes - how AI is revolutionizing cybersecurity

  • 2.
    Meet your speakers Dr. EmmanuelTsukerman Guest speaker Infosec Skills author
  • 3.
    Agenda ● What isML? ○ It automates and scales ○ It’s flexible ○ It allows innovative attacks ● Deepfake live demo ● Few more ML use cases ● Career outlook
  • 4.
    What is machinelearning? Short answer #1: ML automates and scales
  • 5.
    Example: ML formalware detection 5
  • 6.
    6 Traditional anti- viruses (AVs)rely on signatures and heuristics • Signatures: Hashes of files, hashes of snippets of code; akin to cataloging • Heuristics: Similarity of source code to known malicious samples, self-replication, overwriting files, suspicious network traffic, etc.
  • 7.
    7 Malware is toonumerous and evolves too fast >350,000 new malware created every day (https://www.av-test.org/en/statistics/malware/)
  • 8.
    How do Iknow traditional AVs don’t keep up? 8
  • 9.
  • 10.
    ML-based AVs arenow industry standard 10
  • 11.
    What is machinelearning? Short answer #2: Flexibility
  • 12.
  • 13.
  • 14.
    Problem: There isa never-ending array of methods of attack 14
  • 15.
    Illustration of threatdiversity 15 DOS: SMURF, NEPTUNE, Back, land, PoD, Teardrop, TCP SYN flooding …
  • 16.
    16 Illustration of threat diversityfrom firewalls ● Packet fragmentation: Allows hackers to evade pattern matching by fragmenting their packets ● Spoofing source IP address: Hackers appear like legitimate users to the firewall ● Spoofing source port: Allows hackers to bypass port-specific rules in the firewall New attacks developed everyday
  • 17.
    17 🙶 It’s humanlyimpossible to keep up with this ever- evolving threat landscape 🙶
  • 18.
    ML can keepup with and catch: ● Zero-day attacks ● Anomalous behavior 18
  • 19.
    What is machinelearning? Short answer #3: Innovative attacks
  • 20.
    20 You won’t believewhat Obama says in this video! —BuzzFeedVideo
  • 21.
    21 This type ofcrime is already here “Fraudsters Used AI to Mimic CEO’s Voice in Unusual Cybercrime Case” —Wall Street Journal
  • 22.
  • 23.
  • 24.
    Hacking CAPTCHA systems 24 CAPTCHAis meant to tell humans and computers apart, but we can break it using ML
  • 25.
    Smart fuzzing 25 ● Fuzzingis about finding inputs to a program that will cause it to break ● The best fuzzers rely on ML (e.g., genetic algorithms)
  • 26.
  • 27.
    Plenty of greatjobs 27 The primary profession for applying data science to cybersecurity is “cybersecurity data scientist”
  • 28.
    28 ML is beginningto touch all aspects of cybersecurity Areas influenced: ● Malware ● Intrusion ● Social engineering ● Pentesting ● Data security Best learn it, because the bad guys are
  • 29.
    Dr. Tsukerman’s Infosec Skillscourses infosecinstitute.com/authors/emmanuel-tsukerman/
  • 30.
  • 31.
    Everyone gets afree week of Infosec Skills. Then it’s just $34/month infosecinstitute.com/skills
  • 32.
    About us At Infosec,we believe knowledge is the most powerful tool in the fight against cybercrime. We provide the best certification and skills development training for IT and security professionals, as well as employee security awareness training and phishing simulations. infosecinstitute.com 708.689.0131 32