Direct sequence spread spectrum transmissions (DSSS) are now widely used for secure communications, as well as for multiple access. Since the transmission uses a large bandwidth, the power spectral density of a DS-SS signal can be below the noise level. Hence, such a signal is difficult to detect. In this paper, we propose a method which is able to detect a spread spectrum signal hidden in the noise. The method does not require a priori knowledge about the spreading sequence used by the transmitter. It is based on two parallel computations: the “theoretical path”, in which we compute the theoretical behavior of the fluctuations of second order moments estimators in the case noise alone is present, and the “experimental path”, in which we compute the actual fluctuations. When a DS-SS signal is hidden in the noise, the results provided by both paths diverge, hence the presence of the signal is detected. Experimental results show that the method can detect a signal far below the noise level.