The four-stage arbiter PUF is the most popular among the common PUFs. Unfortunately, approximately 40,0 0 0 CRPs are required to conduct ML attacks on a four-stage PUF [3] . To increase PUF security and stability and defend against ML attacks, a novel TSMCA PUF is proposed. A secure PUF structure must satisfy the avalanche effect. That is, as one bit (or several bits) of the challenge changes, more than half bits of the response flip. An illustration of the TSMCA PUF structure with limited resources is shown in Fig. 3 . The PUF structure includes a four-way selector between the arbiter and the multiplexer, in which the selector chooses the path of the signal for the arbiter, and the response must be sent to the arbiter via the selector. The selector is controlled by five bits, so 24 paths are made available for selection each time, thus achieving high reconfigurability. Compared with the four-stage arbiter PUF, the proposed structure requires fewer hardware resources and a smaller area. The challenged order of both PUFs is opposite, and the attacker needs to record two challenge information at the same time, which can effectively increase the difficulties of the attacker attacking the PUF.