A signal scrambling based joint blind channel estimation, activity detection, and data decoding (SS-JCAD) scheme is proposed for coded massive random access systems. The signal scrambling, which is a symbol-wise phase rotation to each user’s modulated data, has no bandwidth expansion and energy consumption. Yet, with it, each user’s channel state information (CSI), including both the user’s active status and channel coefficient, is blindly detected by a low-complexity single-user maximum a posteriori estimation without the help of pilots. With the estimated CSI as a prior, a soft parallel interference cancellation based multi-user decoding is performed. A soft information iteration algorithm between the channel estimation and multi-user decoding is developed to improve the reliability of both. Simulation shows that for massive random access systems with system load less than 1.5, SS-JCAD achieves almost the same bit error rate as the ideal case that perfect CSI is available, implying SS-JCAD as a near-optimal solution to the random access problem.