Tracking of moving objects is crucial to security systems and networks. Given a graph $G$, terminal vertices $s$ and $t$, and an integer $k$, the \textsc{Tracking Paths} problem asks whether there exists at most $k$ vertices, which if marked as trackers, would ensure that the sequence of trackers encountered in each s-t path is unique. It is known that the problem is NP-hard and admits a kernel (reducible to an equivalent instance) with $\mathcal{O}(k^6)$ vertices and $\mathcal{O}(k^7)$ edges, when parameterized by the size of the output (tracking set) $k$ [5]. An interesting question that remains open is whether the existing kernel can be improved. In this paper we answer this affirmatively: (i) For general graphs, we show the existence of a kernel of size $\mathcal{O}(k^2)$, (ii) For planar graphs, we improve this further by giving a kernel of size $\mathcal{O}(k)$. In addition, we also show that finding a tracking set of size at most $n-k$ for a graph on $n$ vertices is hard for the parameterized complexity class W[1], when parameterized by $k$.