1: \begin{abstract}

2: 		\noindent Motivated by TRACE algorithm~\cite{curtis2017trust}, we propose  a trust region algorithm for finding second order stationary points of a linearly constrained non-convex optimization problem. We show the convergence of the proposed algorithm to ($\epsilon_g, \epsilon_H$)-second order stationary points in   $\widetilde{\mathcal{O}}\left(\max\left\{\epsilon_g^{-3/2}, \epsilon_H^{-3}\right\}\right)$ iterations. This iteration complexity is  achieved for general linearly constrained optimization without cubic regularization of the objective function. 

3: 	\end{abstract}