Please enable JavaScript.
Coggle requires JavaScript to display documents.
Data Security and Privacy (Protection (Access control (Example (Bell…
Data Security and Privacy
Threats
Random corruption
Software flaws
Human errors
Malicious corruption
Malicious injection
Protection
Access control
Mandatory
Admin controls all r/w/x permissions
Discretionary
Each user decides
Example
Bell-LaPadula Model
Multi-level security
1) A lower security subject cannot read a higher security object
2) A higher security subject cannot write to a lower security object
Confidentiality
Biba Integrity Model
1) A higher security subject cannot read from a lower security object
2) A lower security subject cannot write to a higher security object
Integrity
High-water Bell-LaPadula
After higher security subject writes to lower security object, increase security level of object to level of subject
Low-water Biba Integrity
After higher security subject reads from lower security object,
decrease security level of subject o level of object
File access control
Access control matrix
Access control list
Which subjects can read/write/execute this object
Capabilities
Which objects can this subject read/write/execute
Biometrics
Visual
Sound
Fingerprint
Gait
Code
Token devices
Error checking/correction
Detection
Append a tag to a file
Checksums
e.g. CRC32
Hashes
A weak cryptographic hash may be a good error detection hash (e.g. MD5)
Input can be any size, output is fixed
(32-bit for CRC32, 128-bit for MD5)
Parity
Correction
Error Correction Codes(ECC)
Hamming code
Backup
Full backup
Differential backup
Stores all changes between current time and last full backup
Incremental backup
Physical security
Storing data
Adobe breach(2013): Passwords were encrypted, not hashed
Replication
Synchronous
Asynchronous
Data Privacy
Sensitive attributes & Personally Identifiable Information(PII)
Dossier effect
Restrict queries
Inference attack
Queries only including Bob
Difference of queries
Intersection of queries
Methods
k-anonymity
Each set of identifiers in the table
must appear at least k times
(= anonymity sets have at least k elements)
Quasi-identifiers
Complementary release attack
l-diversity
Each anonymity set must have at least l different attributes
Anonymization function is usually deterministic
Differential privcay
Anonymization function is random
iOS 10 (2016)
Secure multiparty computation
Private information retrieval