This paper studies an attack-defense game with incomplete information. The game involves two players: the defender and the attacker. The defender locks objects. The attacker then tests the objects and obtains information with random errors. When attacking an unlocked object, the attacker wins the game. Sonin?s complex ?fill-and-switch? strategy is reduced to a simple threshold rule: the attacker targets nodes with a negative signal when the sum of sensitivity and specificity is greater than one, nodes with a positive signal when the sum is less than one, and randomizes when the sum is one. The defender allocates locks uniformly at random. Exact expressions for the attacker?s equilibrium strategy are obtained, as well as analytical expressions for the game value. The game value satisfies duality, reaches its minimum when the sum of sensitivity and specificity is one, and converges to the node value at rate O( 1 n). This makes the LBT model a practical tool for security applications.