BLP

Overview

The Bell–LaPadula (BLP) model (1973) enforces confidentiality via security levels: subjects have clearances, objects have classifications. Two rules: no read up (cannot read higher-level data), no write down (cannot write to lower-level data).

Model

[request_definition]
r = sub, sub_level, obj, obj_level, act

[policy_definition]
p = sub, obj, act

[role_definition]
g = _, _

[policy_effect]
e = some(where (p.eft == allow))

[matchers]
m = (r.act == "read" && r.sub_level >= r.obj_level) || (r.act == "write" && r.sub_level <= r.obj_level)

Policy

You usually need no p rules; the matcher implements BLP from request levels:

No Read Up: Subjects cannot read objects with higher security levels
No Write Down: Subjects cannot write to objects with lower security levels

Examples

Request Examples

alice, 3, data1, 1, read    # alice (level 3) reads data1 (level 1) - ALLOWED
bob, 2, data2, 2, read      # bob (level 2) reads data2 (level 2) - ALLOWED
charlie, 1, data1, 1, read  # charlie (level 1) reads data1 (level 1) - ALLOWED
bob, 2, data3, 3, read      # bob (level 2) reads data3 (level 3) - DENIED (No Read Up)
charlie, 1, data2, 2, read  # charlie (level 1) reads data2 (level 2) - DENIED (No Read Up)

alice, 3, data3, 3, write   # alice (level 3) writes data3 (level 3) - ALLOWED
bob, 2, data3, 3, write     # bob (level 2) writes data3 (level 3) - ALLOWED
charlie, 1, data2, 2, write # charlie (level 1) writes data2 (level 2) - ALLOWED
alice, 3, data1, 1, write   # alice (level 3) writes data1 (level 1) - DENIED (No Write Down)
bob, 2, data1, 1, write     # bob (level 2) writes data1 (level 1) - DENIED (No Write Down)