Typically, mobile application assessments myopically test only the software living on the device. However, the code deployed on the device, the corporate web services backing the device and any third party supporting services must be “whole-isticly” tested AS WELL AS testing the interactions between these components to reach an acceptable level of software assurance for mobile applications.
Mobile Application Assessment By the Numbers: a Whole-istic View
1. SESSION ID:
Mobile Application Assessment By The
Numbers – A Whole-istic View
MBS-F02
Dan Cornell
CTO
Denim Group
@danielcornell
2. #RSAC
Agenda
u Background
u Mobile Application Threat Model
u Assessment Methodology
u Data Collected
u Findings
u Types of Vulnerabilities Identified
u Where Vulnerabilities Were Identified
u How Vulnerabilities Were Identified
2
4. #RSAC
Introduction
u Data comes from:
u 61 Assessments
u 20 Applications
u What we found:
u 957 Vulnerabilities
u Assessment with the most vulnerabilities: 3 assessments had 10 Critical vulnerabilities
u Assessments with the least vulnerabilities: only three assessments had one
vulnerability (all others had more)
4
5. #RSAC
Research Background
u Mobile application threat model
u Assessment methodology
u Static versus dynamic testing
u Automated versus manual testing
u Why CWE?
u Assessment data
5
6. #RSAC
Mobile Application Threat Model
u More complicated than a “typical”
web application threat model
u Not just about code running on the
device
u Main components:
u Mobile application
u Enterprise web services
u 3rd party web services
6
7. #RSAC
Assessment Methodology
u Testing activities
u Combination of both static and dynamic activities
u Combination of automated tools, manual review of automated test results and manual testing
u Tools include Fortify SCA, IBM Rational AppScan, Portswigger BurpSuite
u Scope can include:
u Code running on the device itself
u Enterprise services
u 3rd party supporting services
7
8. #RSAC
Determining Severity
Based on customized DREAD model
u Damage potential
u Reproducibility
u Exploitability
u Affected users
u Discoverability
u Each factor ranked 1-3
Collapsed to single dimension
u Critical: > 2.6
u High: 2.3 – 2.6
u Medium: 2.0 – 2.3
u Low: < 2
8
9. #RSAC
Why CWE?
u Vulnerability taxonomy used was MITRE’s Common Weakness
Enumeration (CWE)
u http://cwe.mitre.org/
u Every tool has its own “spin” on naming vulnerabilities
u OWASP Top 10 / WASC 24 are helpful but not comprehensive
u CWE is exhaustive (though a bit sprawling at times)
u Reasonably well-adopted standard
u Many tools have mappings to CWE for their results
9
10. #RSAC
Assessment Data
u Subset of mobile assessments
u Mostly customer-facing applications from financial services
organizations
u Primarily iOS and Android applications
u Some WAP, Windows Phone 7
10
12. #RSAC
Types of Vulnerabilities Found
u Top 10 Most Prevalent CWEs – Overall
u Top 10 Most Prevalent CWEs – Critical/High Risk
12
13. #RSAC
Top 10 Most Prevalent CWEs – Overall
13
14
14
16
20
21
21
22
26
271
284
0
50
100
150
200
250
300
Use of a Broken or Risky Cryptographic Algorithm - LOW RISK
Information Exposure Through an Error Message - LOW RISK
Cross-Site Request Forgery (CSRF) - LOW RISK
Information Leak Through Debug Information - LOW RISK
External Control of System or Configuration Setting - LOW RISK
Improper Input Validation - LOW RISK
Improper Sanitization of Special Elements used in an SQL Command ('SQL Injection') - CRITICAL
Cleartext Transmission of Sensitive Information - LOW RISK
Information Exposure - LOW RISK
Information Leak Through Log Files - LOW RISK
14. #RSAC
Top 10 Most Prevalent CWEs – Critical/High Risk
14
1
1
2
3
3
3
4
6
6
22
0
50
100
150
200
250
300
Uncontrolled Resource Consumption ('Resource Exhaustion') - CRITICAL
Failure to Preserve Web Page Structure ('Cross-Site Scripting') - CRITICAL
Missing XML Validation - CRITICAL
Uncontrolled Resource Consumption ('Resource Exhaustion') - CRITICAL
Incorrect User Management - CRITICAL
Exposure of Access Control List Files to an Unauthorized Control Sphere - CRITICAL
Access Control (Authorization) Issues - CRITICAL
Access Control Bypass Through User-Controlled Key - CRITICAL
Information Leak Through Caching - HIGH
Improper Sanitization of Special Elements used in an SQL Command ('SQL Injection') - CRITICAL
15. #RSAC
OWASP Top 10 Mobile Risks
u Similar to the OWASP Top 10 Web Application Risks, but targeted at
mobile applications (obviously)
u Top risks to mobile applications:
u https://www.owasp.org/index.php/
OWASP_Mobile_Security_Project#tab=Top_Ten_Mobile_Risks
u Work in progress to update this based on industry-contributed data
15
16. #RSAC
OWASP Top 10 Mobile Risks
M1: Insecure Data Storage
M2: Weak Server Side Controls
M3: Insufficient Transport Layer
Protection
M4: Client Side Injection
M5: Poor Authorization and
Authentication
M6: Improper Session Handling
M7: Security Decisions Via Untrusted
Inputs
M8: Side Channel Data Leakage
M9: Broken Cryptography
M10: Sensitive Information
Disclosure
16
17. #RSAC
Compare to OWASP Top 10 Mobile Risks
17
Strong Overlap
• Weak server-side controls
• Poor authentication and
authorization
• Security decisions via
untrusted inputs
• Sensitive information
disclosure
Overlap
• Insecure data storage
• Insufficient transport layer
data protection
• Improper session handling
• Side channel data leakage
• Broken cryptography
Weak Overlap
• Client-side injection
18. #RSAC
Where Did We Find Overall Vulnerabilities?
18
Corporate Web
Service
591
62%
Device
342
36%
Third-Party Web
Service
24
2%
19. #RSAC
Where Did We Find Critical/High Risk Vulnerabilities?
19
Corporate
Web Service
41
70%
Device
15
25%
ThirdParty
Web Service
3
5%
20. #RSAC
Analysis of “Where” Data
u Mobile security is about more than
the code running on the device
u The things we really care about
(Critical, High) are most frequently
found on corporate web services
u Then on the device
u Then on 3rd party web services
u Reflects the “scale” benefits of
finding web services vulnerabilities
20
21. #RSAC
How Did We Find Vulnerabilities?
u Static vs. dynamic testing
u Automated vs. manual testing
u What techniques identified the most vulnerabilities?
u What techniques identified the most serious vulnerabilities?
21
22. #RSAC
Static vs. Dynamic Method of Finding Vulnerabilities
22
Critical, 10
Critical, 33
High Risk, 14
High Risk, 2
Medium Risk, 84
Medium Risk, 9
Low Risk, 206
Low Risk, 599
0
100
200
300
400
500
600
700
Dynamic
Static
23. #RSAC
Static vs. Dynamic Method of Finding Vulnerabilities
23
Critical
5%
High Risk
0%
Medium
Risk
2%
Low Risk
93%
Static
Critical
3%
High Risk
4%
Medium
Risk
27%
Low Risk
66%
Dynamic
24. #RSAC
Critical and High Risk Vulnerabilities
u Static testing was more effective
when finding serious (Critical and
High) vulnerabilities
u But it also found a lot of lower-risk
vulnerabilities (as well as results
that had to be filtered out)
24
Found with
Dynamic
Testing
24
41%
Found with
Static
Testing
35
59%
Critical/High Risk Vulnerabilities Found
25. #RSAC
Automated vs. Manual Method
of Finding Vulnerabilities
25
Critical, 33
Critical, 10
High Risk, 1
High Risk, 15
Medium Risk, 4
Medium Risk, 89
Low Risk, 526
Low Risk, 279
0
100
200
300
400
500
600
Automatic
Manual
26. #RSAC
Automated vs. Manual Method of Finding
Vulnerabilities
26
Critical
6%
High Risk
0%
Medium
Risk
1%
Low Risk
93%
Automatic
Critical
2%
High Risk
4%Medium
Risk
23%
Low Risk
71%
Manual
27. #RSAC
Automated vs. Manual Method of Finding
Vulnerabilities (Critical and High)
u Automated testing was more
effective when finding serious
(Critical and High) vulnerabilities
27
Found with
Automated
Testing
34
58%
Found with
Manual
Testing
25
42%
Critical/High Risk Vulnerabilities Found
28. #RSAC
Automated vs. Manual, Static vs. Dynamic Methods
28
Cri.cal,
33
Cri.cal,
10
Cri.cal,
0
High Risk, 1
High Risk, 14
High Risk, 1
Medium Risk, 4
Medium Risk, 84
Medium Risk, 73
Low Risk, 526
Low Risk, 206
Low Risk, 5
0
100
200
300
400
500
600
Automatic / Static
Manual / Dynamic
Manual / Static
Automa.c
/
Sta.c
Manual
/
Dynamic
Manual
/
Sta.c
Low
Risk
526
206
5
Medium
Risk
4
84
73
High
Risk
1
14
1
Cri.cal
33
10
0
30. #RSAC
Automated vs. Manual, Static vs. Dynamic for
Critical and High Vulnerabilities
30
Automatic, 34
Automatic, 0
Manual, 1
Manual, 24
0
5
10
15
20
25
30
35
40
Static
Dynamic
Static Dynamic
Manual 1 24
Automatic 34 0
31. #RSAC
Analysis of “How” Data
u A comprehensive mobile application security assessment program
must incorporate a significant manual testing component
u Automated tools for testing mobile applications are not as mature as
those for testing web applications
u Web services can be challenging to test in an automated manner
31
32. #RSAC
On-Device Vulnerabilities By Platform
Platforms Number of
Assessments
on Device
Number of Total
Vulnerabilities
on Device
Average Number of
Vulnerabilities Found per
Assessment
iOS 39 252 6.5
Android 19 84 4.4
Windows Phone 7 1 3 3
WAP 1 3 3
32
33. #RSAC
Other Observations
u We also include “other observations” as part of our assessments
u These reflect:
u Application weaknesses
u Coding flaws or behavior that are not “best practice” but do not reflect an
immediate, exploitable vulnerability
u We had 1,948 “other observations”
u Roughly twice as many as actual vulnerabilities
33
34. #RSAC
Other Observations – Where Were They Found?
34
Corporate
Web Service
55
3%
Device
1892
97%
Third-Party
Web Service
1
0%
35. #RSAC
What Does This Mean?
u Most of these “other observations” are about code on the device
u Mobile application developers need help building better code
u AND automated code scanning tools need to be better about filtering less
valuable results
u Something that is not a problem today could be later on
u Identification of new platform vulnerabilities
u Changes coming along with a new application release
35
36. #RSAC
Conclusions
u What To Test?
u Mobile “apps” are not standalone applications
u They are systems of applications
u Serious vulnerabilities can exist in any system component
u How To Test?
u Mobile application testing does benefit from automation
u Manual review and testing is required to find the most serious issues
u A combination of static and dynamic testing is required for coverage
36
37. #RSAC
Recommendations
u Plan your mobile application assessment strategy with coverage in mind
u Evaluate the value of automation for your testing
u More “cost” than simply licensing – deployment time and results culling
u Look for opportunities to streamline
u Fast application release cycles can require frequent assessments
u Control scope:
u Assess application changes (versus entire applications)
u Manage cost of reporting
37
38. #RSAC
Next Steps (For Us)
u Incorporate more assessment data
u Possible collaboration with OWASP Top 10 Mobile Risks
u Currently being reworked based on data sets such as ours
u Better analysis of applications over time
38