Cybersecurity: It`s All About The Assumptions

Cybersecurity:
It’s All About The Assumptions
Mahalingam Ramkumar
Dept. of Computer Science and Engineering
Mississippi State University
Protecting Digital Assets
Active vs Passive
Why is passive better?
Why are things the way they are?
What needs to be done?
Protecting Assets
Active
Passive
Protecting Digital Assets (Cybersecurity)
Active
Passive
Active vs Passive
Active approaches: elimination/removal, isolation of
Undesired functionality
Undesired traffic
Passive approaches: Check if “everything is OK”
Tools: Cryptography, System Models
Active: Based on attacker model
Passive: Based on system model
Security in a Nutshell
Assumptions
Strategy
Desired
Assurances
Desired assurances: Do not depend on the approach
Security Protocol: Strategy to convert reliable assumptions to desired assurances
Assumptions: The guard is trustworthy; the team that developed the IDS is trustworthy,
the hash function is preimage resistant, etc
Strategy: Is the strategy correct? Integrity of execution of the strategy?
Also boils down to assumptions...
Good Assumptions
Good assumptions should be universally verifiable
The enemy knows the system (Shannon)
The system is completely open (Kerckhoff’s Principle)
Complexity is the enemy of Security
Assumptions behind active approaches can not meet such requirements
Passive approaches can be based on good assumptions
Desired Assurances and Efficacy
Unlimited freedom for attacks: What are the specific goals? All that active
approaches can eliminate is some attacks (even if integrity of execution is
assured).
Passive approaches see any system as a set of data items, with unambiguous
rules for reading/modifying data-items. If the agency for execution (for checking
enforcement of rules) is trustworthy, desired assurances can be realized.
Ideally both approaches should be used: active approaches to prevent as many
attacks as possible; passive approaches to detect effects of those attacks that
had slipped past active approaches.
Toolbox For Passive Approaches
Cryptography
System Models
Cryptography:
Utility: ability to amplify trust
Trust in integrity/privacy of a small number of bits can be amplified to provide
integrity/privacy assurances to any number of bits.
Integrity: hash function based protocols
Privacy: encryption/decryption based protocols
Cryptographic Protocols
Constructed by suitably chaining cryptographic primitives (block ciphers and hash
functions)
Secrecy protocols (for bulk and random access) constructed by chaining block
ciphers
Integrity protocols (for bulk and random access) constructed by chaining hash
functions
For most purposes the primitives are interchangeable
Cryptography Assumptions
Small number of bits stored in a trusted boundary
Correctness of cryptographic protocols (which have read/write access to
protected bits) executed within a trusted boundary
Standards
Trust in the integrity of cryptographic primitives stems from meticulous
standardization efforts
Such a meticulous standard is also required for the environment for execution of
cryptographic protocols
Without such a standard we simply can not make use of the full power of
cryptography.
System Models
Example: Clark-Wilson system integrity model
Data items whose integrity needs to be assured: constrained data items (CDIs)
Only well regulated transformation procedures (TPs) can modify CDIs
Execution of TPs due to external triggers: unconstrained data items (UDIs)
Evolution of the system can be seen as a sequence of snapshots of CDIs
Assumptions for Model Based Approaches
Correctness of model
Integrity of execution environment for TPs
Integrity of environment where snapshots and code representing TPs are stored
Joint Standard Execution Environment
For both cryptographic protocols and TPs
Snapshots of CDIs and TP code can be protected using cryptographic protocols
All we need is for the standard environment to also execute TPs
Even for complex systems TPs can be simple
Complex TPs can also be broken down into multiple simpler TPs
Trusted Tasks
Depending on trigger fetch appropriate TP (verified against protected bits)
Each TP will modify one or a small number of CDIs (fetch them and verify against
protected bits)
If any CDI is private decrypt using protected bits
Execute TP, re-encrypt private CDIs
Modify CDIs, and hence protected bits
Standard environment requires only simple functionality
Beneficial Side Effects
Passive approaches reduce motivation for attacks
If passive methods are also used, assumptions behind active approaches need
not be universally verifiable.
It is enough if the deployer is convinced of its utility
The Bias Towards Active Approaches
Active approaches are more challenging (hard problems); so more appealing to
researchers
Tools used in active approaches are valuable for students (sought after skills in
the industry)
Promoting Passive Approaches
Rigorous standard for execution environment is a prerequisite for reliable passive
approaches
Different skill sets for practitioners of active vs passive approaches
Attacker model (active approaches) vs system model (passive)
Practical deployment issues vs design considerations
Summary
Passive approaches are backed by better assumptions
Passive approaches can substantially strengthen active approaches
Currently passive approaches are virtually nonexistent (research, teaching, and
practice)
Requires completely different skill sets compared to active
A standard operating environment could go a long way (for making full use of
passive approaches)