Type confusion can occur when processing a H264 packet. In the method PacketBuffer::FindFrames in modules/video_coding/packet_buffer.cc there is a loop on line 296 that goes through the data_buffer_ vector backwards. The flag is_h264 is set before this loop, and if it is true, the loop extracts and sets h264 struct specific data in each packet of the buffer. This flag is not updated for each packet. So if a number of non-h264 packets are followed by a h264 packet, a VP8 or VP9 packet can be treated at a h264 check, allowing several bounds checks to be bypassed.