The vanishing short integer solution (vSIS) assumption [Cini-Lai-Malavolta, Crypto’23], at its simplest form, asserts the hardness of finding a polynomial with short coefficients which vanishes at a given random point. While vSIS has proven to be useful in applications such as succinct arguments, not much is known about its theoretical hardness. Furthermore, without the ability to generate a hard instance together with a trapdoor, the applicability of vSIS is significantly limited. We revisit the vSIS assumption focusing on the univariate single-point constant-degree setting, which can be seen as a generalisation of the (search) NTRU problem. In such a setting, we show that the vSIS problem is as hard as finding the shortest vector in certain ideal lattices. We also show how to generate a random vSIS instance together with a trapdoor, under the (decision) NTRU assumption. Interestingly, a vSIS trapdoor allows to sample polynomials of short coefficients which evaluate to any given value at the public point. By exploiting the multiplicativity of the polynomial ring, we use vSIS trapdoors to build a new homomorphic signature scheme for low-degree polynomials.