Targeting passive attackers, a new security protocol based on the inherent security of linear network coding is proposed. Our scheme does not employ any auxiliary transformation or cryptographic routine at any network entity. Moreover, it avoids imposing any throughput reduction. The probabilistic weak security is achieved by selecting "proper" encoding vectors for the outgoing channels at every network node. The probability of finding proper encoding vectors and the probability of protocol security assurance drastically increase as the code field size or the number of transmittable symbols increase, or as the attacker becomes more limited in accessing independent channels.