To improve the security and stability of biometric handwriting samples a Bio-Hash algorithm for handwriting was introduced in [1]. It utilizes features to describe how the sample was written, but the current set of features does not characterize the visual appearance of the sample itself. In this paper we present a set of new features derived from handwriting forensics and OCR algorithms to address this issue. Furthermore, here the security of the old and new sets of features is evaluated for their resilience against a new, fully automated attack trying to compute raw data matching a given hash vector.The main contributions of this paper are: The introduction of new features with a potential to increase the attack resilience of the Bio-Hash algorithm, and, an improvement of the attack approach from [6] to produce more realistic looking synthetic handwriting signals.