Friction in vehicle suspensions is an important property for modeling vertical dynamics. Conventionally, suspension friction is modeled with a simple Dahl friction model in full vehicle simulation models. However, this approach appears to be insufficient for investigations involving degraded shock absorbers or semi-active damping systems. At high shock absorber excitation velocities, the suspension friction is overshadowed by the fluid force of the shock absorbers. When investigating the effect of degraded shock absorbers due to oil and gas loss on the driving dynamics of passenger cars, the correct representation of suspension friction is particularly important, since the masking effect of the hydraulic force is absent for these conditions. This work therefore presents a phenomenological mathematical model of suspension friction, which takes into account the dependencies of suspension friction on the deflection velocity of the vehicle shock absorbers, the oil level of the vehicle shock absorbers and the horizontal forces on the wheel carriers. To parameterize the model, measurements of a MacPherson axle and a fixed beam axle as well as of various vehicle shock absorbers are analyzed. To validate the model approach, a test vehicle with identical axles is used to perform single obstacle crossings at 8 km/h and sinusoidal steering at 50 km/h for the intact vehicle and for oil-empty front axle shock absorbers and oil-empty rear axle shock absorbers. The driving tests are simulated using a mathematical, characteristic curve-based five-mass full vehicle model. The simulation results show that the presented friction model is necessary to obtain a high level of accordance between the simulation and experimental investigation of driving maneuvers with degraded shock absorbers. If the suspension friction for the vehicle with oil-empty rear axle shock absorbers is modeled using a Dahl friction model considering only nominal conditions, that is neglecting the aforementioned impacts on the amount of friction, a root-mean-square error that is 65 times higher is calculated.