Message dropping by intermediate nodes as well as RF jamming attack have been studied widely in ad hoc networks. In this paper, we thoroughly investigate a new type of attack in intelligent transportation systems (ITS) defined as Beacon Non-Transmission (BNT) attack in which attacker is not an intermediate vehicle, but rather a source vehicle. In BNT attack, a vehicle suppresses the transmissions of its own periodic beacon packets to get rid of the automated driving misbehavior detection protocols running in ITS, or to mount a Denial-of-Service (DoS) attack to cripple the traffic management functionality of ITS. Considering BNT attack as a critical security threat to ITS, we propose two novel and lightweight techniques to detect it. Our first technique bases its detection by assuming a certain distribution of the number of beacons lost from a vehicle while accounting for loss due to channel-error. However, it fails to classify shortish BNT attacks wherein amount of denial and channel-error loss are comparable. Our second technique, suitable for identifying both shortish and longish BNT attacks, considers beacon loss pattern of a vehicle as a time-series data and employs autocorrelation function (ACF) to determine the existence of an attack. In order to trade-off detection accuracy for equitable use of limited computational resources, we propose a random inspection model in which the detection algorithm is executed at random time instances and for randomly selected set of vehicles. We have performed extensive simulations to evaluate the performance of proposed detection algorithms under random inspection and a practical attacker model. The results obtained corroborate the lightweight nature of both techniques, and the efficacy of ACF based technique over simple threshold based technique in terms of higher detection accuracy as well as smaller reaction delay.