27
AI in Security Operations: What We’ve Learned So Far Matt Walmsley, EMEA Director [email protected]

AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

AI in Security Operations: What We’ve Learned So Far

Matt Walmsley, EMEA Director

[email protected]

Page 2: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

1. What has AI learned from security operations?

2. What has security operations learned from AI?

3. What does security operations look like after AI happens?

Questions for today

Page 3: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

1. What has AI learned from security operations?

Questions for today

Page 4: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Why do you need a SOC?

Two things we can do in security operations are:

1. Slow attackers down

2. Accelerate detection and response

TIME

An effective SOC provides the benefit of speed of response time to a security incident

Page 5: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Cost

Impact

Initial

infection

Time

Mo

ney

Time to

contain

Time to

detect

Data Loss, Exfiltration,

Intellectual Property

Consulting, Incident Response,

Damage Control

Legal,

HR, Fines

Time is money for security operations

Average cost of a data breach is $3.92 million1

Total cost of Maersk ransomware attack = $350 million

1 2019 Ponemon Cost of Data Breach Study

Page 6: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Data Loss, Exfiltration,

Intellectual Property

Consulting, Incident Response,

Damage Control

Cost

Impact

Time

Mo

ney

AI accelerates detection & response

Initial

infection

Legal,

HR, FinesReducing time to detect and contain significantly lowers cost

Time to

contain

Time to

detect

Page 7: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Dwell Time + Contain Time

=

Unauthorised Access Time to Asset

Where is time spent?

Page 8: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Threat awareness

• Knowledge of:

• Threat presence

• Intent

• Activities

Agility

• The rapidity of which identified intrusions are successfully:

• Contained

• Irradiated

• Normative business operations restored

Security operations dimensions

James Webb – CISO;

Appalachian State University

Page 9: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Incident response maturity

James Webb – CISO;

Appalachian State University

Agility

Th

reat

aw

are

ness

1

2

3

5

Maturity Typical

Detection

Typical

Response

Risk

Awareness

5. Predictive

Defense

Internal

(Hunting,

Waylay) +

External

Highly proactive Very High

4. Intelligence

Driven

Internal

(Hunting) +

External

Threat /

Adversary

Driven

High

3. Process

Driven

Internal

(Hunting) +

External

Service driven

(SLAs)Medium

2. Tool Driven /

Signature BasedExternal Tool Driven Low

1. Reactive /

Adhoc

External,

User Report

Reformat,

Reinstall,

Restore

Very Low

4

Page 10: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

2. What has security operations learned from AI?

Questions for today

Page 11: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

The five questions that data science answers

Is this A or B (or C or D)? Classification

How much / How many? Regression

How is this data organised? Clustering

Is this weird? Anomaly

What action should be taken? Reinforcement

Page 12: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

TTPs /

Methods

Tools

Network

Artifacts

Domain

IP

Looking for what the threat doesD

ete

cti

on

Du

rab

ilit

y

Transient

Persistent

Threat CoverageKnown Novel

Hash

Signatures

Threat Intel

AI Models

Fast, labeled coverage of known threats

• Tools

• Exploits

• Known attacker infrastructure

• Environment-specific indicators

Durable coverage

• Both novel and known attacks

• Difficult and expensive to evade

Page 13: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Combine data science and security research

Security Research

• Identify, prioritize, and characterize fundamental attacker behaviors

• Validate models

Data Science

• Determine best approach to identify behavior

• Develop and tune models

Attacker Behavior models

• High-fidelity detection of things attackers must do

• No signatures: find known and unknown

Page 14: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

It’s all about detecting attacker behaviors

Command and Control

Advanced C2: human control

Botnet C2

Reconnaissance

Network sweeps and scans

Advanced: AD, RPC, shares

Lateral Movement

Stolen accounts

Exploits

Backdoors

Exfiltration

Data movement

Methods, e.g. tunnels

Security Research

• Identify, prioritise, and characterise fundamental attacker behaviours

• Validate models

Data Science

• Determine best approach to identify behaviour

• Develop and tune models

Page 15: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Prioritising incidents using a time series of events

Botnet monetization

Internal recon

Lateral movement

Acquire dataExfiltrate

data

Standard C&C

Custom C&C

Custom C&C & RAT

Opportunistic threats

Targeted threatsInitial infection

Page 16: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

3. What does security operations look like after AI happens?

Questions for today

Page 17: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

• When to automate

• When not to automate

• Keep humans in the loop, don’t give the all keys to the machines

• Free up analysts time for higher value, higher interest work

Automation that augments and empowers

Page 18: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

SOC Visibility Triad: Operational Flow

Visibility, Detection, Analytics Correlation Coordination, Response

Page 19: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

SOC Visibility Triad: Operational Flow

ENDPOINT

NETWORK

Visibility, Detection, Analytics Correlation Coordination, Response

Page 20: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

SOC Visibility Triad: Operational Flow

ENDPOINT

NETWORK

Endpoint

Network

Users

Applications

Threats

Vulnerabilities

Users, Applications, TIP, etc.

SIEM

PRIORITIZED

INCIDENTS

Visibility, Detection, Analytics Correlation Coordination, Response

Page 21: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

SOC Visibility Triad: Operational Flow

ENDPOINT

NETWORK

Endpoint

Network

Users

Applications

Threats

Vulnerabilities

Users, Applications, TIP, etc.

OR

CH

ES

TR

AT

ION

an

d A

UT

OM

AT

ION

ENDPOINT

NETWORK

USERS,

APPLICATIONS

(AD, IDM, CASB,

etc.)

SIEM

Firewall

VERIFIED

INCIDENTS

Network

Alerts

Endpoint

Alerts

Visibility, Detection, Analytics Correlation Coordination, Response

Page 22: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Vectra Cognito: The ultimate network detection and response platform

Cognito platform

Investigate and hunt in a cloud-

based application

Cognito Recall

Send security-enriched metadata

to data lakes and/or SIEM

Cognito Stream

Detect and prioritize hidden

threats at speed using AI

Cognito Detect

• Cloud, user, datacenter

• Security-enriched

• Real time and historical

• Scalable architecture

Page 23: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Workload reduction from triaging, correlating and prioritising events into incidents

Source: https://www.vectra.ai/download/2019-black-hat-edition-of-the-attacker-behavior-industry-report

Workload reduction of 174X

14,644 / (57+27) = 174.33

“This solution excels at rolling up numerous alerts to create a single incident

to investigate that describes a chain of related activities, rather than isolated

alerts that an analyst then has to piece together.”

- Gartner Market Guide for IDPS

Page 24: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

AI delivers real-world security operations benefits

“Vectra Cognito helps close the skills gap"

“Vectra Cognito makes threat hunting more efficient”

”With Cognito, I can focus on the highest-risk threats”

“Vectra has helped us reduce our business risk”

Page 25: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Commodity trading business

7000 employees, 60 countries, $10bn revenue

15,000 customers, 14,000 supplier

By self admission at level 1-2 security maturity

Looking at a security programme which was “Security by Policy”

2 people to run this programme

Entered a POC for detection alongside EDR

Automated and integrated SOC quickly delivered an efficient and powerful service

Off-loaded policy to a single person who built policy from industry regulations and SOC interactions

SOC now at level 4-5 security maturity

Saved significant amounts of money for their company and has had increase in budget based on performance

SOC visibility in action

Vectra Cognito AI powered Network

Detection and Response (NDR) Platform

Download case study

Page 26: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

1. What has AI learned from security operations?

2. What has security operations learned from AI?

3. What does security operations look like after AI happens?

Time is the most important metric for detecting and responding before damage occurs

Machine learning works best when applied to a specific task

Increased threat awareness and operational agility

Answers for today

Page 27: AI in Security Operations: What We’ve Learned So Far · 1. What has AI learned from security operations? 2. What has security operations learned from AI? 3. What does security operations

Thank you

vectra.ai